2nd Meeting - 2nd Session Group of Governmental Experts on Lethal Autonomous Weapons Systems 2025
Meetings & Events
Date: 1 September 2025
Language: English
Transcript: https://transcripts.un.org/es/asset/k1e/k1eiq3sfis?lang=en
Transcripts available through this tool are created by using automatic speech recognition and are not official records nor official documents of the United Nations. Official records and official documents are available on the Official Document System of the United Nations.

---

Chair [0:00]:

Excellencies, distinguished delegates, could I ask you to take your seats again so we can start. I hope you all had a good lunch and again, still some appetite to continue the discussion in the afternoon. We have a number of requests for the floor. We have first the Asia Pacific Institute for Law and Security, to be followed by Bulgaria, Sweden, Article 36, Panama, Norway, University of Nottingham, and Israel. That is what I have on the list now. If there are any further, if you are a bit at the back, please wave your plaque because we can't see it from here, unfortunately. But let me now first pass the floor to the Asia Pacific Institute for Law and Security. You have the floor.


APILS [1:14]:

Thank you, Chair. Good afternoon, colleagues. The Asia Pacific Institute for Law and Security, or APILS, is a specialist organization that seeks to further national and international law as a means for achieving sustainable regional and international security, including human security. which support and promote responsible practices in the development and use of technology for military and security purposes. Chair, we thank you for your continued leadership during the intercessional period, which has led to the revised rolling text that we have before us today. We note the substantive progress that has been made on this text under the current mandate of the GGE. This confirms the wisdom of your approach, Chair, to structuring and organizing the work of the group. The rolling text is a rich document which provides a good set of building blocks for an instrument. We appreciate the opportunity to refine the text further and to make a few comments on box one. Paragraph one offers a working definition of laws for the purposes of these discussions and as an element of an instrument. As we understand it, the intent of this paragraph is twofold. On the one hand, it facilitates our work by setting some boundaries on what we are talking about. On the other hand, it makes it clear that this understanding does not prejudice discussions in other fora which may well adopt different characterizations. In our view, the opening words of paragraph one within the scope of application of the CCW do not best capture that intent. The phrase, "the scope of application" has a specific legal meaning under Articles 1 and 2 of the convention, essentially meaning situations of armed conflict. The phrase is used in the rolling text as a boundary marker in a different sense and for a different purpose. We therefore suggest replacing the first clause of this paragraph with the words, "in the context of the objectives and purposes of the CCW." This phrase comes directly from the mandate of the group. It also has the benefit of firmly anchoring the elements of the instrument in the CCW's role of prohibiting and restricting the use of weapons that are excessively injurious or have indiscriminate effects. We see merits in the US proposal of bringing the introductory clause of paragraph one along with the opening clause of paragraph two to the very beginning of box one as a chapeau. This would then make it clearly applicable to all three paragraphs in this box. Moving on to the phrase, an integrated combination of one or more weapons and technological components. Our colleagues from Article 36 have previously pointed out that the words "integrated combination" could imply that all functions must be performed by a single physical unit. We agree that any risk of such a reading should be avoided. In practice, laws are likely to be made-up of physically dispersed components which operate as a system to achieve autonomous functionality. Indeed, there already exist point defense systems which are modular, consisting of physically separate weapons, sensor units, and control units. To better reflect the fact that we are dealing with weapon systems, but not necessarily unitary devices, we could perhaps use the phrase interconnected combination, or alternatively, functionally integrated combination. We also listened with interest to the proposal of the Russian delegation to add the word "auxiliary" before the words "technological components." Language of this nature could indeed help clarify the role of those additional technological components and thereby address some of the concerns that were raised by the Indian delegation earlier. Finally, on paragraph 1, we appreciate the omission of target identification from the list of critical functions of laws, which has been supported by many delegations this morning. And we found the arguments offered by the ICRC before lunch especially persuasive. Many of the concerns raised in this GGE apply even when a human identifies a set of targets. but the weapon system needs no further human input for detecting these targets and selecting and engaging one of them. The measures of this group in develop-- the measures this group is developing in boxes three and four of the rolling text would be perfectly relevant to such weapon systems where target identification is performed by a human. Paragraph 2 of Box 1 seeks to clarify the meaning of lethal for the purposes of paragraph 1. We agree with those delegations that prefer not to limit the elements of the instrument by reference to lethality. And these delegations have already explained in detail that the lethality of a weapon can depend on its use and the existing IHL framework makes no distinction between weapons based on their lethality. That having been said, paragraph two represents a constructive attempt at a compromise solution. But we find the second sentence slightly confusing. Grammatically, the sentence suggests that it is the circumstances that result or not result in death. While circumstances are certainly relevant, the focus should be on the use, the human conduct that IHL regulates. We would therefore suggest replacing the words "used in circumstances that" with the words "used in a way that". Also, we agree with Australia that the list of examples of non-lethal harm is unnecessary and could be deleted. With these changes, the second sentence of paragraph 2 could read, quote, "The fact that a laws can be used in a way that does not result in death, does not exclude it from this characterization." End quote. And, Chair, we are very happy to provide our proposed text in full to you by e-mail as well. I thank you, Chair.


Chair [8:16]:

Let me thank you for your very in-depth analysis of the text in front. As has been said before, there is no attempt with regard to integrated discrimination that it should be in one. The object is to make clear that it could be spread out over diverse systems. Any text solutions that would make that clearer are always helpful, and we look forward to receiving your suggestions in writing, as before. Let me have a look at my list. Next on my list is Bulgaria to be followed by Sweden and Article 36. But first, the distinguished representative of Bulgaria, you have the floor.


Bulgaria [9:07]:

Thank you, Chair. At the outset, I would like to express gratitude to you and your team for the creative leadership and devoted efforts towards a successful conduct of the deliberations during the second week of the GGE. Be assured of Bulgaria's full support for you and your team. Bulgaria welcomes the holding of several rounds of informal consultations during the intercessional period, aiming to encourage discussions on the latest version of the rolling text, which factors in important aspects from the March session and offers a well-balanced language. We believe the intercessional period was used in a productive way for coordinating among delegations to research convergences and differences. We also highly appreciate the distribution of the background paper. With regard to box 1, we consider it to be well organized and to contain broad enough working characterization. As pointed out in your background paper, the identification function can be covered by the selection function, so we can support the exclusion of the term "identifying" from para 1. Regarding para 2, in the spirit of compromise, we can support the current formulation, as the concept of "lethality" has been of a quite controversial nature within the GGE. We recognize that little refers to the capacity of a weapon system, which can also encompass damage and destruction activities. Thank you.


Chair [10:53]:

Let me thank the distinguished representative of Bulgaria for stating position of his delegation and making proposals. Next on my list is Sweden, to be followed by Article 36 and Panama. But first, now, the distinguished representative of Sweden, you have the floor.


Sweden [11:11]:

Thank you, Chair, and good afternoon, colleagues. First of all, I would like to thank you, Chair, and your team for all the work you have done already, and you can count on Sweden to support you in your role as Chair in this meeting. With regard to box 1, we are shortly going to go through it, because most of the issues have already been dealt with and given good arguments, but we really like the wording of para 1 as it is currently framed by you, Chair. With regard to para 2, we don't really see the need for it, but we understand that for some parties this is quite important. So in order to be constructive and flexible, we can accept it as it currently stands. Lastly, I would like to echo for those who pleaded for not trying to include too many new words, phrases or changes. Currently, we are looking for elements to be forwarded to our future negotiations. The current wordings are not perfect, but we believe that they are mature and broad enough to enable everybody to engage actively in future negotiations. Hence, we believe we should avoid trying to chisel out certain detailed issues that can and will be addressed later on when we get to those detailed negotiations on the future element. I thank you, Chair.


Chair [12:36]:

I thank the distinguished representative of Sweden for making suggestions and an appeal with regard to the way forward. Next on my list is Article 36, to be followed by Panama and Norway. But first, Article 6. You have the floor.


Article 36 [12:56]:

Thank you, Chair, and good afternoon, everyone. And it seems very good that we're straight back into a continuation of, I think, the very productive, substantive discussion that we that we had earlier in the year. I wanted to comment on a couple of issues within this section, but particularly on the issue of identification, the proposal to add in identify as a necessary characteristic within the broad characterization of lethal autonomous weapon systems. I think as others have said, we're concerned that adding in the term identify as a necessary characteristic could be interpreted in a way that excessively limits the scope of the instrument and excessively limits the scope of the issue of the whole in a way that's a little bit out of step with the approach to regulation that we would see as being taken in the wider text. I guess in particular, the concern is that it could imply that situations, uses of weapon systems where the human user, the human operator has identified the target of attack in advance are somehow then excluded from falling within the purview and the effect of this regulation. And that seems like a very broad loophole to introduce. It does feel like potentially some of this is just coming from an uncertainty about the the meaning of terms that is perhaps inherent in this debate. I think the food for thought paper that you and your team put out earlier in the year was quite useful for identifying that as a concern. I want to suggest a few scenarios that, to me, get at some of the concerns in this area. I am thinking about these scenarios in which it is the same system that is being used in each. If I imagine one scenario in which a human user, commander, operator is seeking to attack a specific enemy armored vehicle that they're aware of on the other side of a hill, well, they may identify that as the target and they may use the autonomous weapon system to travel to the other side of the hill, locate it, and specifically engage it. In another scenario, in a second scenario, using that same system, the human user operator could be aware of a group of armoured vehicles in that same location on the other side of the hill. They use the same autonomous weapon system and it specifically identifies one of those armoured vehicles to engage and to strike. And in a third scenario, the human users are aware of the possibility of the presence of armoured fighting vehicles in that location, but they don't know for certain that they're there and they use that same system to proceed to that area and it engages an armoured fighting vehicle in that location. I guess in these examples, the sort of physical, technical process at work in the system is exactly the same in all cases. What's different is the level of the human user's specific understanding of the target and the military objective that they're operating against. And I think our concern would be that in perhaps the latter of those scenarios, perhaps that wouldn't be considered a scenario that, perhaps that would be the only scenario that would be considered to fall within the scope of the weapon system identifying the target. I guess you could consider a fourth scenario in which a user of a weapon system has no concept of a target or a military objective that they're operating against at all. But if they're using a system with no concept of what they're using it against, then it seems very difficult to see how that would fall within a sort of legal, lawful, responsible use of a weapon system. So If we think about our approaches, if we take select and engage to imply that identification is already occurring to some extent within the system, an interpretation that a few delegations this morning have already articulated, then that would seem to be compatible with treating all of those scenarios, all of which are using the same weapon system, as something that should fall within the requirements for responsible behavior that we're trying to articulate and compile within this instrument. But if we add in that requirement for identification as an additional specific necessary characteristic, it allows an interpretation, doesn't necessarily necessitate it, but it allows an interpretation that would exclude a significant number of those scenarios from being considered treated by the instrument that we're describing here. So in the sort of context of the rules regulatory approach that states are compiling within the wider body of the instrument, it feels like sticking to select and engage provides the broad sense of characterization to which that broad array of subsequent regulations most appropriately apply. Just on a couple of other points briefly, would agree with others that we've always found lethal to be rather awkward in this context and We would, of course, in a way, prefer that it isn't used. But recognizing divergence of opinions in the room, it feels like you've worked your way towards a proposed solution here that's, you know, in some way, hopefully solving the problem for a sufficient number. I guess a little bit wary that with this being necessarily an awkward tension, it might not be possible to wordsmith away to a solution here that fully resolves all of the tensions for all parties, but rather this may necessarily be an area where we have to all be a little bit dissatisfied with what's on the table. And finally, just wanted to comment on the description of an integrated combination of one or more weapons, technological components, appreciating comments earlier from Asia Pacific Institute and some earlier in the day and from your, yourself, Chair, um, in response to some of those comments that our concern here is really to retain a sense that we don't have to be talking only about sort of physically unified, um, system, sort of physically unified objects, but rather we're talking about the connection of functions within a system and recognizing that what we're talking about here is an array of functions within a system rather than necessarily single unified pieces of hardware, I think is important for understanding what we're characterizing in this discussion. Thank you, Chair.


Chair [19:31]:

Thank you for your very developed view on the present box. Next on my list is Panama, to be followed by Norway and the University of Nottingham. But first, Panama. Madam, you have the floor.


Panama [19:48]:

Thank you very much, Chairperson. Given that this is the first time that my delegation is taking the floor, I would like to join others in expressing their gratitude And congratulations to you and through you to your team and the chairperson's friends for the very laudable work you have done. You can count on our support to carry our deliberations to a safe harbor. Regarding the rolling text, it is our view that in its current state, it is the result of a broad, transparent, inclusive consultation process with contributions from states and civil society organizations. The document has reached a sufficient enough level of maturity for us to deem the mandate given to us, the GGE, as satisfied. That mandate was to consider and to draft a set of elements of an instrument advancing its work on a set of elements of an instrument. That said, it could continue to be a foundation for future work to be decided during the next review conference with a mandate that we hope will be to negotiate Protocol VI of the CCW. Regarding box one in paragraph one, we agree with the explanations provided by the ICRC and other countries that have spoken along the same lines. That is to say, we wish to keep the reference to select and engage. In paragraph two, though my delegation had always opted for removing the adjective lethal It is our opinion that the text as proposed by the Chair is a constructive compromise solution, considering the various proposals that have been made over the course of previous sessions of the GGE, given that, on the one hand, it keeps the reference to capacity to cause death which was a point highlighted by several delegations, while also addressing that it can cause injury, damage, or destroy objects. And that had been supported by several delegations, including mine, which is why we support paragraph 2 as presented by the Chair. Thank you.


Chair [22:40]:

I would like to thank the distinguished representative of Panama for stating the views of her delegation and underlining her priorities. Next on my list is Norway, to be followed by University of Nottingham and Israel. But first, let me pass the floor to the distinguished representative of Norway. You have the floor.


Norway [23:03]:

Thank you, Chair, uh, and I'd like to start by joining the chorus, and, and, uh, thank you very much for the hard work that you and your excellent team have done to set the table for us this week. Uh, your paper on context appropriate human control, or judgment and control, and the critical functions, as well as the, uh, informal consultations were very helpful to us. And on a general note as well, we'd like to express our appreciation for the updated version of the roading text, which is an improvement in many regards and remains, in our view, a very solid basis for our further work and future negotiations. But as Australia and many others have mentioned, we still have time left to further refine it and look forward to using that time productively. And on that note, over to box one. In paragraph one, we support the deletion of the word "identify" for the reasons that we stated in March. In short, we want to avoid a too narrow definition. That's been said a couple of times in this room today. And certainly in one of the widespread interpretations of identification, most, if not all, autonomous weapons systems will perform some kind of identification of the targets that they then select for engagement and then go on to attack. But as you highlighted in your paper, Chair, there seem to be differing interpretations of the term. And if you have a cumulative definition, such uncertainties at the terminological level might make the characterization easy to circumvent by claiming, for instance, that the system does not possess the ability to identify the target autonomously. So what we want to do is to ensure that the characterization is inclusive also of autonomous weapon systems that do not have only the most sophisticated identification abilities. And like others, we believe that the selection function can be seen to capture the function of identification. especially when the word selection is not preceded by the word identification in the sentence. If conversely, identification is mentioned, it necessarily leads to a narrower interpretation of selection in order to distinguish the two concepts, which in our opinion is not necessary. And here I think it's also useful to repeat our clear understanding that the characterization itself does not regulate anything. It doesn't impose any obligations or prohibitions. It simply defines the scope of what we are addressing in this prospective instrument, or perhaps more precisely, in our prospective negotiations of the instrument. How we want to deal with the various subsets of weapon systems that fall under this characterization is something that we determine in the paragraphs that follow. And ultimately, of course, this characterization is something that we could and should return to once we start seeing the final shape and contents of our instrument. It's a working characterization mostly, we think, to shape our thinking when we get to the rules and regulations part of our job. With respect to the debate on human user, we support the text as it is, but we can also consider alternative formulations. What is important for us is that we retain the focus on the human element or the absence of it. But if that is expressed as intervention by human user or simply by a human or otherwise, that doesn't seem to make much difference to us. And the proposal to replace human with-- human user with operator, could work for us, but we because because we see like a normal interpretation of the operator is that it is a human that is operating the system. But to remove any doubt, we believe that it should be expressed clearly by including the word human. But we took good note of the arguments that the US made on this point and wonder if a solution could be to use a few more words and write instead, so, "by the human that operates the system." So, like this, "without intervention in the execution of these tasks by the human that operates the system." As regards paragraph 2, we think this is an improvement over the version that you circulated at the end of our March session, Chair, and can support this. And a couple of comments extra at the end. One, the Chinese delegation raised a question regarding how delegations in this room interpret the word "intervention" in the phrase "without intervention by a human user." And we understand that to mean that once the human has activated the system, and the system can select and engage targets all on its own without the operator having to do anything more at all. On the issue of whether to include the word design in paragraph 1, we took note of the Indian argument that if we only focus on the capacity to cause death, we risk characterizing everyday items as weapons. I think we see this a bit differently, and the key for us is in the context and the term "laws" itself, which makes it clear, at least to us, that we are talking about things that can be described as weapons in the normal sense of the word and not everyday items that can be used to kill. And just one point for clarification: the Russian delegation mentioned that in their interpretation of paragraph 1, autonomy is something that the autonomous system has throughout its life cycle. And just to be sure that I understand this correctly, does this mean that the weapon system must have autonomous capabilities other than in the critical functions of selection and engagement in order to qualify as laws, so that means earlier in its life cycle as well? Or does it simply mean that a weapon system that is designed to have autonomous capabilities in selecting and engaging targets is covered by the characterization of laws throughout its life cycle, for instance, also when it is on the shelf and not being used, and not just in the very short phase at the end where the weapon system is actually undertaking these tasks autonomously? Thank you.


Chair [28:38]:

Let me thank the distinguished representative of Norway and also for raising some pertinent questions to further discussion. Next on my list is the University of Nottingham. You have the floor, madam.


University of Nottingham [28:51]:

Thank you, Chair. My observations and recommendations are on the legal interpretation and consequences of the current rolling text. In box one, paragraph 1, the reference to lethal autonomous weapons system is actually not consistent with an understanding in IHL as to what constitutes a weapon. If you look at article 36, Additional Protocol 1, there's a reference to weapon, means and methods, and that's for a very particular reason. Something that may not constitute a weapon will constitute a means. For example, it could be a weapon, or it could be a weapon system, or it could be a platform. In any case, it would still be subject to an Article 36 weapons review. I think to be in line with the understanding under IHL as to what we're dealing with here, whether it's a weapon or a weapon system, I would suggest that there be inclusion in the text of weapon or weapon system. The second point is in relation to paragraph two of box one. I think it's slightly misguided in terms of the perhaps insistence of some to include the word lethal. This actually is misaligned with international humanitarian law, and it creates ambiguity. International humanitarian law is concerned with the nature and effects of a weapon or a weapon system, and those elements are precisely what are being regulated. If we think back to 2016, states parties at that point were still developing their understanding as to what laws are. It is a remnant of that that some may feel the need to have the word "lethal" included. Nevertheless, the proposal simply misaligns with existing IHL, and I recommend that there be rephrasing to refer to the weapon or weapon system that is capable of causing either injury or death, of persons or damage to or destruction of objects. This would align with international humanitarian law, make it very clear that we're dealing with injury, death or destruction to persons or objects. The third point is in relation to what's currently in paragraph one referred to select and engage a target. And I take note of the distinguished delegates from Japan and Korea who eloquently expressed the rationale for including identification, which I would agree with. Identification is an important process in the critical functions. Why? Because it may be through human programming, it may be through advanced machine learning, or it may be a combination of those two. In order to capture that that aspect in the critical function is being regulated under IHL, I think it necessitates to refer to identification here, to make it very clear that this would be falling under IHL itself. And again, I refer to the statements made by the distinguished delegates from Japan and Korea on this point. And finally, distinguished delegate from the United States referred to perhaps replacing the word "user" in paragraph 1 with the word "operator." I think that adds more ambiguity, because it suggests that the system is not autonomous, that there is some interaction with a human, and therefore we could perhaps be dealing with a semi-autonomous system. Whereas what could resolve this issue is if we completely removed by a human user and after without we simply say human intervention so it would read without human intervention in the execution of these tasks thank you Mr. Chair.


Chair [33:36]:

Let me thank the distinguished representative of the University of Nottingham for giving a views on the text in box one next on my list is Israel to be followed by Malawi and Cuba. But first, let me pass the floor to the distinguished representative of Israel. You have the floor.


Israel [33:58]:

Thank you so much, Chair. At the outset, please allow me to thank you, Ambassador Lindenbosch, as well as your team and the Secretariat for your hard work, and to express our delegation's full cooperation and support in the discharge of your duties. Before containing Israel's comments on the rolling text, we wish to express our appreciation for the recent informal meetings, which we found very helpful in clarifying the core issues of disagreement and in moving us closer toward a shared understanding and potential consensus. Concerning the rolling text in box 1. Without expressing a definitive or exhaustive position on the matter of characterization, we have several comments. We agree with the US comment that the beginning of paragraph 2 is also relevant to paragraph 1, therefore we suggest as well moving the phrase "for the purpose of these elements of an instrument" to paragraph 1 and turning paragraph 2 into a subparagraph of paragraph 1. Regarding the first paragraph in box 1 of the rolling text, first, similarly to the proposal by India, we will request to add the words "their integral" before "technological components". This change is necessary to clarify that we are referring to a definable weapon system and not to external components of general purpose which are not actually part of the system. As previously noted, the current text does not sufficiently reflect the fact that the weapon system often rely on a large amount of information and external features that are not inherently related to it. The way it is phrased now makes it very difficult to determine what is considered a part of the system and what not. In addition, like several other delegations, we believe it is important that the word "can" in paragraph 1 be changed to "design" in line with the common terminology in the international law applicable to weapons, including the CCW. We are concerned that the current wording is setting a negative precedent that the legality of weapons and their definition is determined based on the potential use or outcome instead of their design. Our third point concerns the functions of laws. The system's ability to identify a target by itself is an essential characteristic of laws. The GGE has already found consensus on this issue. Systems that only hone in and advance on a target that someone identified are not the purpose of this GGE. We will therefore read any definition as being limited to system that also identify targets but still think it will be useful to clarify in this text by adding "identify" before "select". We think this will relieve some confusion we saw in the June's informal meetings. Finally, like many other delegations, Israel's position is that characterization of laws must focus on lethal autonomous weapon systems in accordance to our current mandate. We believe that adding the phrase "with lethal force" as proposed by other delegations, conveys this idea without reopening the border and complex discussion on the definition of lethality within the Forum. Considering the above, we propose the following wording for the paragraph: "A lethal autonomous weapons system can be characterized as an integrated combination of one or more weapons and their integral technology components that is designed to identify, select and engage a target with lethal force without intervention by a human user in the execution of these tasks. Finally, regarding the second paragraph in box 1 of the rolling text. The previous meetings have shown that the meaning of the term "lethal" is subject to significant disagreements, and the attempts to define lethality are challenging. We are concerned that this issue could divert the forum from reaching consensus. Furthermore, a definition of lethality might have broader implications that extend beyond the scope of our discussion. For these reasons, and from a practical standpoint, we suggest deleting this paragraph and moving the phrase "with little force" to paragraph 1, as we just proposed. To finish, if the definition of lethality remains in the text, we believe it is important to use the words "a weapon system designed to" instead of the words "the capacity to". As mentioned before, IHL rules that regulate weapons, including in the CCW framework, typically differentiate between weapons on the basis of their design and not their hypothetical use. Indeed, almost any means of warfare and even physical objects which were not meant to be weapons can hypothetically be used to cause death or injury. Therefore, the crucial element is the weapon's design. Thank you so much, Chairperson.


Chair [39:56]:

Let me thank the distinguished representative of Israel for sharing his views and making proposals with regard to the present text. Next on my list is Malawi, to be followed by Cuba, Ukraine and Belgium. But let me first pass the floor to the distinguished representative of Malawi. You have the floor.


Malawi [40:17]:

Thank you so much, Chairperson, for giving my delegation the floor. First, we'd like to appreciate the work you have done so far and appreciate the structured rolling text. Like other delegations, we believe that it is a good text to form the basis for negotiations of a legally binding international instrument that promotes peace, human dignity, accountability, and the responsible use of emerging technologies. We emphasize on a legally binding instrument as we are guided by the common African position on laws, which calls for international legally binding instrument that includes prohibition and regulation on laws to uphold international humanitarian law, international human rights law, and the principles of humanity. That being said, on the text in box one, We support the attempt to provide a working characterization of laws. We believe that any system capable of selecting and engaging targets without meaningful control should fall within the scope of regulation. And we support the current wording in paragraph 1 in its entirety. In paragraph 2, we appreciate the attempt to define lethality. However, we believe that it should not only focus on death of a person, but we should also include injuries. So therefore, we'd like to propose that it should read like, for the purpose of these elements of an instrument, lethal refers to the capacity to cause injury and death to a person. Thank you so much.


Chair [42:15]:

I thank the distinguished representative of Malawi for stating the position of her delegation. Next on my list is Cuba, to be followed by Ukraine and Belgium. Let me now pass the floor to the distinguished representative of Cuba. You have the floor.


Cuba [42:32]:

Thank you, Chairperson. To begin, I would like to join other delegations in expressing their recognition and acknowledgement of the work that the Chairperson has done in developing the rolling text, which will be used as a foundation for discussions in this session. We appreciate the proposal that has been presented and the efforts made for us to fulfill our mandate as we have considered more intensely the specific proposals to draft measures that take into account existing protocols in the Convention. guarantee significant human control, and strict respect of international humanitarian law in all circumstances. It is our view that arriving at a consensus definition of autonomous weapons is of vital importance in order to make progress in these debates. As we have said in other times, in our opinion, the definition should consider completely autonomous weapons as well as semi-autonomous weapons. From my delegation, fully autonomous weapons are to be understood as weapons that act without human oversight once they are deployed. Also, semi-autonomous weapons should be defined as those that have human control oversight, at least as regards to critical functions such as selection and engagement of targets. However, we are willing to consider and work on a foundation of the proposals that have been presented. We underscore that the increasing convergence about the need to establish a ban and regulations about these systems points to the fact that this should be one of the priorities of the GGE under the current mandate, which is why the prohibition and regulations must be at the very center of our efforts. Our delegation, as is by now traditional, will maintain a constructive spirit, helping us make constructive and substantive progress. Thank you.


Chair [44:24]:

I thank the distinguished representative of Cuba. Next on my list is Ukraine, to be followed by Belgium, Finland, and Peru. But let me first pass the floor to the distinguished representative of Ukraine. Sir, you have the floor.


Ukraine [44:41]:

As this is the first time that the delegation of Ukraine is taking the floor, we would like to avail ourselves of an opportunity to express our gratitude to you, Chair, and to your team for your outstanding efforts in the development of the rolling text. We're also thankful to the delegations who have provided or envisage providing certain working papers, as well as the delegates of all the peace-loving nations present in this room, representatives of the non-government organizations and the academia. The delegation of Ukraine finds the present rolling text to be in good shape and perfectly capable of serving as a solid basis for the development of an instrument regulating the issues related to the lethal autonomous weapon systems. The delegation of Ukraine sees this meeting as an opportunity to complement the discussion with the most up-to-date practical experience from an ongoing armed conflict, of which we shall speak strictly in terms of use in bello without touching upon the issues of use ad bello. Our delegation is fully satisfied with the approach to the rolling text, which strongly advocates for a future instrument based on the principles of international humanitarian law. Ukraine is deeply convinced that rigorous compliance with the principles of law is a beacon which helps navigate in unknown waters. In our opinion, the task of this group is to properly apply the principles of international humanitarian law to a new type of weapons and in a broader sense, to a new way of waging a war. Without prejudice to the overall satisfaction with the rolling text, the delegation of Ukraine would like to underline the need for precise definitions. We see the purpose of this future instrument in explaining how exactly international humanitarian law principles apply to very particular military capabilities. While working here in Geneva, we must think about the people in the field who are going to use those instruments in accordance with the text that we expect to create. So let us not make them second-guess what we meant, and let us make the definitions as precise as possible. And at this point of our discussion, we would like to touch upon the issue of lethality that was previously hotly debated here. It is the position of our delegation that we must both retain and reinforce the notion of lethality in the definition. As for the reason for retaining, the reality of our time is that artificial intelligence is taking an increasingly notable place in our daily lives, including in warfare. To fit the purposes of the future instrument, our definition needs to be precise enough and it must clearly exclude everything that we do not expect to regulate, including military artificial intelligence, which raises a question of what we're actually expecting to regulate. We believe that regulations should only apply to weapons that can cause immediate destruction and death. Failing to mention this could potentially result in numerous non-lethal capabilities being potentially falling under the definition. We doubt that this is the result that we want to achieve. Therefore, the notion of lethality is essential for the definition. Speaking about the reinforcement of that notion, Our delegation deems it necessary to clarify which capabilities are clearly not included in the definition. This methodology follows the general approach used in certain instruments regulating weapons, for example, by the protocol on incendiary weapons to the CCW that mentions the notion that the notion of incendiary weapon does not cover certain objects that may cause burns as a side effect. We believe that we can use the same clarification for the lethal autonomous weapon system. For example, we believe that it should be clarified that the notion of the lethal autonomous weapon systems must not include the decision support systems that are a very distinct capability from an autonomous weapon systems, and this must be clarified in the definition. We invite other delegations to discuss the definition's possible limitations, and with this, let me conclude the statement and thank you.


Chair [49:09]:

Let me thank the distinguished representative of Ukraine for stating the position of his delegation and making proposals and referring also to other protocols. Next on my list is Belgium, to be followed by Finland, Peru, Cote d'Ivoire, New Zealand and the United Arab Emirates. But let me first pass the floor to the distinguished representative of Belgium. Sir, you have the floor.


Belgium [49:38]:

Mr Chairperson, let me start by thanking you, your team and the Friends of the Chair for your tireless efforts, which have carried us to this point. You have my delegation's full support. In our view, the text in front of us represents a comprehensive and balanced proposal that reflects growing convergences between delegations. The present rolling text is, indeed, the culmination of the work conducted over the past decade or so. While there is room to further refine certain elements, in our view, the substance should be retained. That brings me to box 1. We support the proposed language on the critical functions select and engage, as, in our view, the term identify is a subset of select. We want to commend your efforts in drafting your insightful background paper in this regard. On paragraph 2, if the term lethal needs to be retained in the spirit of compromise, We welcome the language to clarify the meaning of lethality in this specific context. In this case, Belgium favours a broad interpretation of the term lethality, which should be read in conjunction with the two-tier approach contained further in the text, distinguishing between those laws that are to be permitted and those that are not. We believe it makes sense to cast the net wide and later distinguish the captured systems according to the two-tier approach. In this spirit, we would oppose adding "with lethal force" in relation to the critical functions in paragraph 1. I thank you, Chair.


Chair [51:26]:

I thank the distinguished representative of Belgium for making suggestions on behalf of his delegation. Next on my list is The distinguished representative from Finland, you have the floor.


Finland [51:44]:

Mr. Chair, since this is the first time Finland is taking the

floor, let me begin by thanking you and your team for your efforts in preparing the current well-structured rolling texts as well as the informal consultations. I would like to express our overall support for your valuable work in guiding us towards the fulfillment of our mandate. Regarding the text in first paragraph, for us, the clearest version would be the original identify, select, and engage. However, we can support the updated first box on definitions as it stands. Select and engage is clearer than the previous version of identify and/or select and engage. where we found the addition of the word "or" to be problematic in relation to the concept of autonomy. To remain flexible, if we go with this current formulation, we should interpret the select function to include the identification phase. We also appreciate Chair's added proposal on the definition of lethality in paragraph 2, which appears to be a well-worded compromise reflecting our last discussion. Currently, the wording of "capacity to cause death" is quite abstract, and we could also support the more precise option of "capacity to kill" proposed by the US delegation. In paragraph 1, we could also support the proposal of replacing "human user" with "operator". I thank you, Chair.


Chair [53:29]:

Let me thank the distinguished representative of Finland for making concrete proposals and referring to proposals made by other delegations. Next on my list is Peru, to be followed by Cote d'Ivoire. But let me first pass the floor to the distinguished representative of Peru.

Sir, you have the floor.


Peru [53:49]:

Thank you very much, Chairperson. Given that this is the first time that Peru is taking the floor at this second meeting of the GGE, I would like to extend to you cordial greetings on behalf of the government of my country. We wish you every success at the outcome of this second meeting. You can count on my country's full support. On the rolling texts, We believe that this is a significant step forward and towards the goal of having a legally binding instrument in the area, and to have this happen in the near future, given the importance of regulating this type of weapons, given the progress and dazzling progress made in development of new technologies. As is known, in 2023, Peru, along with other states, presented in the DG a document called draft protocol six on autonomous weapons systems as part of the CCAC. A document that reflects the Peruvian position and is a tangible, realistic outcome to address the concerns of the international community as regards the development of these autonomous weapons systems. This draft is a foundation which can be updated. Some of its definitions and provisions can be updated based on the result of future deliberations of this group. Regarding the rolling texts, on the first paragraph, I should say that for Peru, autonomous weapons systems refers to weapon systems that include autonomy in the critical functions of selection, engagement, targeting, engaging, and the use of force, applying force for targets. as pointed out in the proposed protocol 6. This definition could include the term identification. That said, as has been said by several delegations, We, too, are flexible regarding the inclusion or not of that term. Regarding the lethality of weapons, we believe that the adjective used, lethal, commonly used for loss, unduly restricts the scope of this concept. It is convenient instead to refer to all autonomous weapons systems, regardless of their lethality. May I illustrate this with example? civilian facilities using autonomous weapon that might cause a discriminated harm is not only contrary to IHL, but is also part of the weapons that we are trying to regulate, which is why we might have preferred removing the word "lethal" as a sign of compromise and display of flexibility. We agree with paragraph 2, which tries to clarify that situation. Thank you.


Chair [56:43]:

I thank the distinguished representative of Peru. Next on my list is Cote d'Ivoire, to be followed by New Zealand and the United Arab Emirates. But let me first pass the floor to the distinguished representative of Cote d'Ivoire, Ms. Vela-Paul.


Côte d’Ivoire [57:00]:

The

Cote d'Ivoire addresses It's congratulations to you and your excellent team for the work you have done and the text you submitted for us to analyze. We welcome the spirit of transparency and consensus seeking which has guided you in your efforts, and we'd like to congratulate you on the major headway we can see in this document. In particular, we would note the explicit confirmation that international humanitarian law fully applies to laws the recognition of the need to have human control and judgment, which is appropriate, and also the precise formulation prohibiting the use of indiscriminate or uncontrollable systems. This is a firm and solid base for us to build a consensus. In a constructive spirit, my delegation wishes to make the following comments. On box 1, my delegation supports the definition as proposed, including a nuanced approach as to the concept of lethality and also an exclusion clause which allows us to adapt to technological progress. Notwithstanding this, we consider it would be expedient for the text to explicitly provide for identification as a human function that cannot be delegated. And this is vital for selection and engagement. It is absolutely vital. It is critical for the protection of human rights. Moreover, my delegation takes note of the proposal of France. to replace human user by human command chain. We believe that this merits particular attention as an approach because it does correspond to IHL. To conclude, Cote d'Ivoire wishes to reaffirm that we are available to pursue this work in an open-minded manner to achieve a balanced outcome in the best interests of our populations. Thank you.


Chair [59:28]:

Let me thank the distinguished representative of Cote d'Ivoire for stating the position of his delegation and making proposals. Next on my list is New Zealand, to be followed by the United Arab Emirates. That's it so far, so if anybody would like still to react to previous interventions, please raise your plaque, because otherwise we are slowly, slowly wrapping up box one and moving on. But let me now first pass the floor to the distinguished representative of New Zealand. Sir, you have the floor.


New Zealand [1:00:04]:

Thank you, Chair, including for the detailed preparatory work that has gone into this session of the GGE, and indeed the fourth session under your chairing. We are pleased with the amount of progress we can see being made through the evolutions in the rolling text. New Zealand believes the current version of the draft of the rolling text represents a decent balance among the many points of view we have heard expressed, including today. We recognize your rolling texts, paras 1 to 3, take on much of the feedback that delegations have provided before today. Like others, we are concerned by understandings that might limit the applicability of these rules to various scenarios. Like, for instance, adding once more the requirement that a laws must also identify a target before it moves on to select and engage the target. Others have pointed out the problems here, including excluding scenarios where a target has been identified separately. We would also be concerned by formulations that serve to redefine agreed IHL. which is precisely one of the reasons why we have been concerned by an insistence to continue referring to lethal autonomous weapons systems. However, it is our belief that your point two goes a long way to addressing these concerns, including to explicitly include damage to objects and the effect of causing injury. Several constructive suggestions were made during this morning's session. But it is our belief that some of the detailed discussions we are having in here will be more appropriately completed later, ideally as part of international negotiations on an agreement. One of the privileges of our current GGE mandate is the opportunity it provides to formulate elements. That offers us a significant amount of latitude and flexibility. Like our Swedish colleague has just reminded us, the current formulations in your rolling text are mature, and we should be careful to avoid over-engineering a perfect text at this stage of the process. I thank you.


Chair [1:02:24]:

I thank the distinguished representative of New Zealand for stating the position of his delegation. Next on my list is the United Arab Emirates, followed by Algeria, Austria, and the ICRC. But first, let me pass to the distinguished representative of the United Arab Emirates.

You have the floor.


United Arab Emirates [1:02:51]:

Thank you, Mr. Chair. Allow me to address you in Arabic. Given that this is the first time that we take the floor, I would like to express our gratitude to you, Your Excellency, for chairing our meeting. And we would like also to thank all the delegations that have submitted observations or delivered statements. And I would like to present to you the observations of the United Arab Emirates regarding the three parts of our discussion. Regarding paragraph one. The condition of effective human control is absent. Any decision pertaining to the use of lethal force should stay in the hands of the human beings, and this is contrary to the positions of countries in the previous meetings. The current drafting opens the way for totally autonomous weapons, which is a problematic issue that has not been subject to consensus and we do not approve its, uh, um, uh, the reference to it without any conditions. Regarding paragraph one as well, we think that that the absence of human control or intervention should not be understood as an absence of intervention by the human beings at later stages, such as selection, engagement, or design. And we should establish a difference between old systems that are programmed according to some conditions and modern systems that are supported by AI. Regarding paragraph two, We respect the opinions of other countries when it comes to identification. and allow me to present the opinion of the UAE regarding identification or paragraph two. This is a superfluous identification. It is not precise. And when we say level weapons, this should mean widespread destruction and not linked only to the legal scope. and there was no link between the lethal dimension and the rules of international humanitarian law, namely when it comes to identification between civilians and military personnel or other considerations. Regarding paragraph three, it is a paragraph that opens the way to many exemptions in the future, and we support such an openness. However, this openness should be linked to very clear criteria between what is exempt and what should be included, and any exemption should be linked to human control. In this paragraph, we would like also to highlight the importance of having an approach that links the weapon system that target human beings and other weapon system that rely partially on the human dimension. Allow me also to add that we have a reservation regarding the term possible exclusion, because this could not be limited to some clear ethical or military criteria. We hope that the future drafting should have clear conditions. Any amendment or any exemptions or exclusions in the future should abide by the standards of international humanitarian law and an agreed framework. My final observation, appropriate human control is a necessary precondition to guarantee observance of international humanitarian law. And this has been confirmed by various GGE meetings. The current text does not clarify whether autonomous traditional weapons systems, such as the aerial defense systems, whether they are included or not. And we propose that we should establish a difference between what is totally autonomous and the traditional weapon system. Finally, we do not reject the current proposal in its totality. However, we ask for expanding its scope so that it can include legitimate uses or other considerations. Thank you.


Chair [1:08:34]:

Representative of the United Arab Emirates for stating the positions of her delegation and making observations. Let me say that, of course, the characterization is not equals not to prohibitions and regulations. That is a different step. We first start to describe what laws are, and then we go into the topic of which laws can be used in accordance with international humanitarian law and which cannot. Next on my list is Algeria, and then I still have also Austria and the ICRC, but let me first pass the floor to the distinguished representative of Algeria. You have the floor, madam.


Algeria [1:09:18]:

Mr. Chair, as this is the first time my delegation takes the floor, we wish to express our appreciation to you for your efforts and your team for preparing the session of the GGE. You can count on my delegation for support. We consider the rolling text a solid basis for the GGE deliberations, including the future possibility of developing a legally binding instrument to regulate laws. We followed carefully the discussion of this morning, and we wish to deliver the following remarks regarding box one of the rolling text. On paragraph 1, we think that it is important to reach a common understanding of the concept of an integrated combination of one or more weapons and technological components. We also concur with the views expressed by many delegations regarding the need for a broad and practical approach in defining laws in this first stage. Regarding paragraph 2, We share the position voiced by many delegations on the importance of identification as a critical stage. This is the phase in which a system detects, recognizes, and classifies potential targets. At this stage, human control must be ensured, particularly to guarantee the accurate identification of military targets and to distinguish them from civilian infrastructure, which is protected under IHL. I thank you, Mr. Chair.


Chair [1:10:59]:

I thank the distinguished representative of Algeria for stating the position of her delegation. Next on my list is Austria, to be followed by the ICSC. But let me first pass the floor to the distinguished representative of Austria. You have the floor.


Austria [1:11:15]:

Thank you very much and apologies for entering a second round, but I think it's very useful and we definitely want to have an interactive discussion in some of the arguments that have been brought forward, but also some formulations need more clarity and we would like to go back to some key points. And in our case, this would be the discussion around identify and the discussion around lethality and to some extent also about the integrated combination. Starting out with identify, I think it's important for all of these three elements to repeat, and I think you have made this point, but also other delegations, is that this is the definition, or at some point we also called it the characterization of laws, and the purpose is to define the scope. So what is the scope of a possible instrument? What would be relevant for our discussions and what would not be relevant for our discussion? A scope by nature must be as broad as possible so that we don't have any loopholes or any blind spots, and it also encompasses things normally that per se would not be regulated in an instrument. So it is normally not related to the rest of a legally binding instrument. It sets out the scene, and what comes later is the prohibition and the regulation. And I think there's still a lot of confusion when it comes to this currently in the room. It is not the place to limit the scope of what we are doing here. It is not the place to make the prohibitions or the regulations part smaller, because otherwise we will end up with a crooked instrument that just will have the effect that additional instruments, additional protocols, or even amended protocols will be necessary, and I think this is not something that we want to have. Second point, I feel that we are having a repetitive discussion here. We are going our third round. Our, our, it's a third, our third round when it comes to identify, select or engage and I wouldn't be surprised if we end up again with identify and/or select in our next round because this is the kind of circular, uh, version of our discussions that we are seeing, seeing, seeing currently. At the same time, and this is maybe my third point, we understand the discussion to some extent as delegations don't want to make the scope of our instrument or about the set of elements of an instrument too broad. There have been made examples, and that is often useful, that we don't want to talk about fire and forget missions, that that would be too extensive. On the other side of the spectrum, we also have the problem that if we are going too narrow, we are having also weapon systems that clearly have a high level of autonomy or that are using artificial intelligence to a large extent that in a general assumption would definitely be autonomous weapon systems. And we will come to that later, what that means. What we have to find is a balance that it's not too broad or not too narrow. Currently, many of the proposals that we are hearing are going too narrow. We are having also a certain risk here on the scope that we don't know how technological developments will go forward. So if we are going to narrow, we will have a lot of systems that we don't know yet that would fall outside our definition, and that is definitely a bigger problem. In general, we see a too broad scope less of a problem than a too limited scope, and this is something that I think you also tried to address in your working paper. So I will just ask delegations to have a look again at your working paper on the different types of systems and how this applies to the scope. Next point on the identify discussion. Several delegations have pointed out that identify is an essential part of a targeting cycle. I would argue that identify is definitely something that is important when it comes to artificial intelligence, and this is also something where delegations have referred to the 2019 report. The 2013 report, in a segment where we talked about definitions and characteristics, talked about the role and impacts of autonomous functions in the identification, and this is an important part of our work where we should further look into. I think everyone would agree with that, that the role that autonomy plays when it comes to identification of targets is a crucial issue of our work. But as many delegations have pointed out, this aspect of identification can be found in the select part that we are using here already. The bigger problem when it comes to identify is that it is a very vague word and that it has very different uses. And as I mentioned, some have pointed to the targeting cycle. If you're looking at just some targeting cycle that you can find in public information, for example, the NATO trying targeting cycle, identification is not one of the main parts of the cycle. It is a subtopic under find. And this is probably the most broad interpretation that you can use when it comes to identify. It is more or less just saying this is a possible target, what do we do about it? And if this is something that includes in a cumulative approach in our definition, we are creating a huge problem and it will definitely damage our work here and it will damage the output of whatever we are doing here. Sorry for being so long, but I think this is important. Other delegations have asked for Examples. And I just want to say again that we appreciate examples of systems that might fall under this definition and are and should not. I think this is important, as the Indian delegation has pointed out, about fire forget, but also about certain ballistic missile systems. This is something we have to keep in mind and look into. Maybe there are opportunities to do it, but to some extent, some of these weapon systems could also be seen as collateral, because by their functioning per se, the way they are used, the way they are designed, they might not fall under this definition as set out in the later part of the text. And I think this is an exercise that we should look more closely into. Examples that might fall outside the scope if we use approach where we only look into systems that are doing the exercising identification, selection, and engagement. It would be, for example, light ring munition that works only on the basis of radar signals because there is no identification at the beginning, it's just a general direction. Light ring munition sent into a predefined area with pre-identified targets would, under this definition, also fall outside of our scope. Autonomous undersea vehicles on a long-term area denial mission to attack certain targets that come into a certain area and that are carrying payloads, but not per se have an identification mission when it comes to their predefined targets. Swarms, which select and distribute targets on the basis of predefined targets as well. especially if there is kind of a collective identification process and then individual elements of the swarms would only do the distribution of targets or the selection of the targets. I think what is in general important is also the space and time that lies between identification and select. If we have a broad interpretation of "identified" and we are not very clear when and how this happens, this could be something that someone is doing months before a military mission and that there is no direct link or even correlation between the identification process and actually the selecting of a specific target. And this is just another gigantic loophole that is created if you go in for a cumulative approach. Sorry. Second part of my intervention about lethality. I think what we are currently discussing on para 2 is not in line with our previous discussion, because the purpose of this para has always been that we make clear that lethal is not a narrowing element of our discussion, that we are using the word lethal in the sense that this is something that is related to the use of force and that could actually do severe damage to civilians and civilian objects, but also to combatants and other protected groups under IHL. Insofar that we are now looking into a way how to narrow the definition of lethality is concerning, because this is not how the original idea by the ICRC to come up with a definition of lethality for the purpose of these elements of an instrument Uh, was intended, and we are actually changing the intention of, of the original, uh, drafting of this para, and, uh, we cannot go along with this, unfortunately. It also has to be mentioned that lethality is not a category of IHL. We are quite surprised that lawyers are so eager to go forward with such an exercise and to make up a definition that would, and someone has mentioned this as well, that would carry on into other discussions when it comes to weapon systems. So if we are going down this path, we will probably create a precedent for future arms control discussions, and it is not necessarily a good one. lethality is an effect of a weapon system and not something embedded in its design. There have been some references to protocols under the CCW where the design element is also included, but there mostly we are talking about the effects of a weapon system. And here this is some of them, and this has been pointed out in the past, are not necessarily lethal. So the design to be lethal is not the way that we can continue. And the previous protocols do not serve as a good example in that regard. What we think is dangerous is if we are going into extremely limited understanding of lethality. It does not only narrow the scope of our discussions, it also goes against the purpose of the CCW, which is mostly about the prevention of indiscriminate effects or superfluous injury. And it, as I mentioned already, creates a dangerous precedent for future discussions. And as Brazil has pointed out, if you're just looking into anti-personal weapon systems, we are actually, after more than 10 years of discussing lethal autonomous weapon systems, we are actually narrowing the scope of our discussions in hindsight, because most of us have not been aware that we are talking about only anti-personal weapon systems. Most of us here in this room had an understanding that we are talking about all autonomous weapon systems. So this is, to some extent, I have to say, unfortunately, also a bad faith effort if we are going too narrow on the lethality aspects. I will close just a quick word on the integrated combination. I think this addition, and that's why we were curious about the understanding of other delegations, what it means. And this was also pointed out by Raint from the civil society. It's a useful concept if it's meant broad, that it tries to, not integrate, but include all kind of weapon systems that are also passed over different locations and systems. But if it means that all of them should be physically located in one place, then it creates a problem. So we definitely want to see this as a broad definition. And we are not in favor of, of a very, uh, limited interpretation of what it means that we are dealing with an integrated combination, uh, I leave it at that and sorry for being so long.


Chair [1:23:36]:

I thank the distinguished representative of Austria also for giving the views of his delegation with regard to statements made by other delegations earlier in the day. I have two more requests for the floor, one from the ICRC and one from the United Kingdom. But let me first pass the floor to the ICRC. You have the floor.


ICRC [1:23:58]:

Thank you, Mr. Chair, and thank you very much to all the delegations who have spoken so far this morning and this afternoon. We found the discussion very illuminating and some very helpful ideas have been put forward. I wanted to react to some of the points that colleagues have raised about select and engage or identify, select and engage, the so-called critical functions. And I'm very pleased that Everyone who's spoken today has really taken your advice and given reasons for their views, and they've explained very clearly why they think it's important to do it this way or that way. It's very helpful to our work. So I'd like to take up two, well, specific examples raised by two delegations, that is, the delegation of India and of Singapore, who both mentioned more or less this, whether weapons that are fire-and-forget missiles or guided missiles would be covered by this characterisation if we just have select and engage and we don't have the addition of identify. For us at the ICRC, it's really very clear and the whole point of what we're trying to do here is determine what happens when it's not a human that is deciding on the specific target to be engaged. So in the case of a missile being fired, even if it's a fire and forget missile that guides itself to the target, it is still the human, the soldier, the pilot, the operator who is selecting that specific target. They select the target and then there's no question what it is that the weapon is going to engage. It may fail of course, it may miss or hit the wrong thing, but it is going to an identified target and it's a responsibility of the operator, of the pilot or whoever's firing that weapon to ensure that IHL obligations are fulfilled. So in those cases those weapons, fire and forget missiles, would not be covered by the characterisation as we have it here, select and engage, because it is not the weapon that is selecting the target autonomously. The target has already been selected. It's already been identified, if you want to put it that way, and selected. So the weapon, even if it's guiding itself in a very sophisticated manner, an automated manner, it is just engaging, it's not selecting. So we think select and engage covers what we need to cover. And I think this whole discussion illustrates really that there's very little disagreement on substance in the room, I think, about what should be and shouldn't be included in the scope. It's just a disagreement about what the appropriate words are to capture that. And with that in mind, I'd like to come back to something a colleague from New Zealand mentioned a few minutes ago, that is we are still, you know, we are just identifying the elements and we don't need to narrow down every, nail down every last detail. That is a task for the negotiations that will follow on an instrument. And so, you know, And in those negotiations on an instrument, for example, we could have quite expansive definitions of what exactly select means, what exactly engage means, that would really clarify. And I'd also like to thank you for your background paper, which gave some very concrete examples of the different kinds of identify and select and so on. And again, I recommend delegations refer to that because it does help to show why we are having trouble with this. And It also helps to show that the disagreement is really just about the wording. It's not about what we think should or should not be included. I think there's a lot of common ground on that. Finally, of course, the long discussion on lethal, to be expected. I think here again, to come back to New Zealand's point, We just need to get an idea of what it is that we're trying to regulate here, and trying to nail down to the last degree what exactly lethal means or doesn't mean is perhaps not the most useful use of our time in this setting under this mandate. We note the proposals to include with lethal effect, the words with lethal effect in paragraph 1. We don't even really want to get into an argument about that, because I think it will just take us back to what we've already gone round and round several times. The proposal you have in the text now, in paragraph two, is an attempt, I think a very good attempt, to bridge the gap between two quite different views on this. ICRC, of course, maintains its view that we should not have the qualifier "lethal" in the text at all. But your proposal to try and reconcile these two positions, I think, is a good one. And if we try and improve it or narrow it down further, I think we'll just go back into the same old circular debate. So again, if that does need to be solved, it's something that should be better solved in the subsequent negotiations on an instrument. Thank you.


Chair [1:28:59]:

I thank the distinguished representative of the ICSC also for reflecting on views presented earlier today by other delegations. Next and last on my list is the United Kingdom. You have the floor.


United Kingdom of Great Britain and Northern Ireland [1:29:16]:

Thank you, Mr. Chair. And as this is the first time that the United Kingdom is taking the floor, I'd like to thank you, your team, and the friends of the Chair for all of the work that has been put in to get us to the very positive place that we are at this moment. And please be reassured that the United Kingdom will continue to assist you in all of this work. Addressing the issue of characterisation, at this particular moment, there is a clear difficulty within this group to include both the terms, identify and select, as it is clear that states understand these terms differently. and this risks obscuring rather than clarifying the scope of the type of system that we are meant to be characterizing. Where there does appear to be agreement is in the need to align these terms with the requirements of IHL. However, the critical issue is not whether or not this group agrees on this requirement, but rather whether the nature of the system in question and its performance of the particular functionality that we are trying to describe in this exercise presents a unique challenge in respect to that performance. Part of the problem here is that the alternative understandings of these terms relevant to the function that we are describing are potentially interchangeable. As suggested in your paper on this point, the manner in which the term "identify" is being used by a number of delegations potentially attracts the qualities that are considered central to the discriminative function, which represents the technological step change from the non-autonomous system to the autonomous system. However, if we look at the question purely definitionally, the term identify is most commonly understood as recognize, classify or name, whereas the term selection is more clearly associated with a choice which is more naturally associated with the IHL concept of discrimination. A further aspect is the point or moment when the question of can the target be engaged in compliance with IHL is answered. If that question is addressed within the identify function, then it would appear to be correct to have the term identify within this characterization. The concern would be that we do not wish to narrow the application of any instrument by suggesting that there are cumulative criteria that need to be satisfied for the measures to apply. Therefore, if the recognition element of the term identify is a necessary precondition of the selection function, then it would be less critical to include it in this characterisation. However, we would note that we are in a multilingual environment The UK has heard the concerns from other delegations that there is not sufficient clarity inherent in their own understanding of those terms to provide a clear basis for a characterisation. There are a couple of ways that the group might address the issue. We could use papers such as your May paper as a body of work to aid in understanding of the meaning of the terms as presented in the rolling text. Or we could find a short additional formulation of words that resolves the questions of whether the term identify is necessary to properly capture critical autonomous functions. The UK position is that we believe that the term select is a clear and unambiguous and sufficient to address the critical functions we wish to capture in this characterization. Thank you, Mr. Chair.


Chair [1:33:04]:

Let me thank the distinguished representative of the United Kingdom for really reflecting on the discussion and in a way already taking away a bit from the points I would like to make. Now we are, I think, at a kind of a natural end or exhaustion of the discussion about box one. Let me try to say a few words looking back at this indeed very rich discussion, sometimes almost academic, and with all due respect to the academia in the room, we should be careful not to become a debating club. We have a clear mandate. I'm a very practical person. At the end, we need to have a text. I think it was very well said by the UK delegate also, we need to get this kind of a common understanding. I have the same feeling like he has that if you listen carefully to what delegations say, there is not that much difference, but it is the way we are phrasing it. And with all due respect to the excellent work done by the translator, some of the points might also be lost in translations because simply it's not always easy to capture it in the right words. So I don't think that we should change the text as we have it now. We should really dive into clarifying what we should be covering by select and engage. And I see possibilities for that. One of the reasons for our paper was, and the discussion we had at the end of May was indeed to already facilitate that. I understand some of you are still working also on some text that might be helpful in the coming days. So I would really stimulate to have this kind in the corridors and around the meeting discussions by experts to clarify that. I think there have been some positive proposals also with regard to integrated combination. There again, I think we know what we mean, or at least I hope that I've been able to make clear what I would like to achieve with the text. If it's then still not clear in English, we have to see what we can do about it, a few suggestions. made, I think, with regard to lethality. We have to be careful not to turn around in circles, but also there have been made a few, I think, helpful suggestions we could work on without totally rehashing the language. Yes, it was said also by one of the delegations that Is there already a decent balance between the views expressed? Well, there's only a decent balance if all delegations recognize that. So that is something we have to establish. It's clear that in the end we will, I said it already at the opening remarks, we will not have perfect English, we will not have the truly, truly perfect definitions. But as long as we agree on what we have on paper, that we have the same intent, that is, that is, of course, what we should achieve also to make things workable in the future. And I think it was also said by the delegate of Austria, be careful with regard to too much discussion on the scope. This is a characterization. And if we, at a moment in the future, with a different chair, have discussions on a possible instrument, and I don't say what character, then we no doubt will have to sort out certain things and be more precise. And depending on the character then of the instrument, it could be wider or less wide. I think that was already said at the first meeting in 2024 by the distinguished ambassador of China. I think that is it for today with regard to box 1. And I'm now looking for my speaking points to move on to box number 2. Can we have it on the screen? Yeah, it fits. because we should not forget about number five, but that is probably the one we are going to pay most of the attention to. In this box, and let me briefly highlight a few of the changes which take into account several of the comments put forward during the March session, we notably made para 4 less law specific by removing at the beginning "involving laws" And most importantly, after careful consideration, I kept context-appropriate human judgment and control in the text. I have by now become very familiar with both support and criticism of this wording. And no doubt we will discuss this in detail again. But I would also like to point out that I think the terminology has something for all. There is some flexibility in context-appropriate. And by including both judgment and control, I feel we have text that could work for most, if not yet all, delegations. Um, and that's it for my side, and I would now really like to hear your views, and I open the floor for comments and suggestions. Who would like to kick off? I recognize the distinguished representative of the Russian Federation. You have the floor.


Russian Federation [1:39:28]:

Thank you, distinguished Chair. Before we move to box 2, I would just like to make one comment still on box 1, in particular to react to the question which was put by the delegation of Norway on those considerations that we put forward as to box 1. In particular, relating to the proposal to delete reference to task 2 can select, identify and target in para 1. to delete the wording in the execution of these tasks. This proposal was called for, we felt, by the fact that our premise is that this function will be in the weapon systems not only when they are fulfilling a particular task, executing the task when they are being used. They will be there in the system from the very start. They can be there in the programming and in the development and the manufacture of these systems. Therefore, these characteristics will be an inalienable part of the system throughout the whole of its life cycle, and not only at the point in time when the system is fulfilling a task to select and engage a target. So that is why. If we say that we have a definition only for those systems which will have autonomy at the point when they are executing these tasks to select and engage a target, it's something we feel would not be justified, would not be correct. As in those cases, as we said, the question can be raised as to the need to include an understanding in the next part of this document, next box, relating to the development, the programming, the manufacture of these systems, and also legal reviews of these weapons systems. As at the stage of design and development of these systems, the function of autonomy, if we look at the working definition that we have now, that function might not exist, as here we are only referring to having autonomous functions that can select and engage a target just in the execution of these tasks. Moreover, we would also like to make a brief comment on para 3. It's our view that here we require more clearly setting out in this paragraph the concept that the working definition, which is in para 1, does not cover specific concrete kinds of weapons. At this point already, we think it's important to do so, as we've discussed the fact that there's a series of proposals from delegations relating to specific categories of weapons which should be excluded from this definition so that we don't discuss certain kinds of weapons in the context of this discussion. Moreover, to have a very clear definition of the fact that some delegations are talking about some instrument which might be developed or perhaps other measures on these weapons systems, it's absolutely critical that we define this here. that we set it out very clearly, otherwise we run the risk of really encountering major difficulties when we seek to agree upon any outcome and conclusions relating to these weapon systems. Turning to box 2, distinguished Chair, on paragraphs 1, 2 and 3, our delegation does not have any particular comments to make. These paragraphs are fully acceptable to us and they are in line with the previous understanding that we achieved in the context of this GGE. And they are in line with the guiding principles which were approved by States party to the Convention. And we very much count on the fact that the formulation of these paragraphs will remain unchanged as We would really insist upon that. Then turning to paragraph 4, the substance of this paragraph does not give rise to any particular comments from us either. We see in this paragraph, we have the fundamental understandings that were agreed upon in the group previously, in previous years. which are the basis for future discussions on this matter and should be taken into account by all States participating in the discussions. But we do have some problems with the initial wording. We did raise these concerns at the previous session of the group and we're prepared to repeat them. At this point, we consider that it is not expedient to make this link with the Convention and its protocols or other international agreements either, as it is our view that this goes somewhat beyond the mandate that we have and creates some grounds for different interpretations in the future, which we would wish to avoid. So instead of the first sentence, we feel it to be justified and to be correct in the context of these discussions to mention the fact that the main part of paragraph 4 refers to lethal autonomous weapons systems. So we could just limit it to relating with respect to lethal autonomous weapons systems, and then carry on with the understanding that we have here in paragraph 4. And so, we would be solving two particular problems. We would move away from referring to the convention and other international agreements, and we would also be making a direct link with lethal autonomous weapons systems, which is what we are focusing on here. And then on para 5, we have two comments to make on para 5. The first comment relates to the concept of human control, human judgment and control. It is our view that the formulation, the wording which has been used to reflect this understanding, to this notion, is not quite correct. Previously, in the activities of this group, we restricted the notion to either control or judgment. And we would prefer, in this particular case, to restrict ourselves just to one of these notions. In this case, we would find it preferable to restrict ourselves to human control as a notion. Most of the States participating are in favour of that. We are prepared to agree with the use of that notion, that concept in paragraph 5. And as we said earlier, in the spirit of flexibility and compromise, we would also We prepared to accept human judgment. But using both of these notions is not something we feel to be particularly necessary as human control as a notion we believe covers the concept of human judgment as human control without any human judgment is not really possible. And then again, human judgment is not the same as human control, and it does have a more narrow meaning as to what comes under this notion. Human judgment is narrower than human control, which is broader in nature. Therefore, In this particular case, using both of these notions, both of these concepts, seems to us to be superfluous and not fully justified. Then, on the context appropriate characterization of this, we would prefer to avoid using such a term. As this would require a definition of this term, and we need to have that definition now, at this stage, we would need to agree on what comes under this term, what exactly is context appropriate, and what is not context appropriate. And we know full well that the understanding of appropriate in itself is rather complex, in particular when we look at legal categories. Given this, we would prefer not to use this term and endeavour to express what we understand as being human control including if we mean taking account of particular contexts when this human control is exercised through a longer, more expressive formulation, so that it is very clear and easy to understand exactly what we mean. And so, once again, we'd repeat it for us. we would find human control acceptable and human judgment acceptable as well. And we would be prepared to use either of them in the recommendations and outcome of this group relating to the particular aspects of lethal autonomous weapons systems. Then, further on in this paragraph, We would prefer to delete the reference to the use and effects of laws. It's our view that we do not need to bring everything under the stage of execution, the use. We believe we should be more logical in the understandings that we have worked on in this group and we'd like to recall that previously in this group we did talk about human control in general, human judgment in general, for the whole of the life cycle of these weapon systems. We consider that that understanding needs to be kept further. In particular, because further up in paragraph 4, we refer particularly to that, to the fact that it is throughout the whole of the time period, at all times, that this weapon system should be under international humanitarian law. Therefore, we will conclude that human control should then be throughout the whole of the life cycle of such systems. Also, we would recall that in further parts of this document, we clearly specify the fact that such control must be ensured throughout all stages of the life cycle. In addition, we understand that in deleting this formulation, we might possibly lose the link with the lethal autonomous weapons systems. And we would not like to lose this link in paragraph 5. Therefore, we consider it necessary to make a reference to lethal autonomous weapons systems in the context of ensuring human control by adding in a sentence relating to lethal autonomous weapons systems after the sentence on human control. So that there's a clear understanding that it is human control relating to lethal autonomous weapons systems. Thank you.


Chair [1:56:59]:

I thank the distinguished representative of the Russian Federation to react to previous statements on the topic of box one and for going into detail with regard to the position of his delegation with regards to box number two, in particular with regard to context appropriate human judgment and control. I look Oh, at least I have one more request so far, and that's from the Republic of Korea. You have the floor.


Republic of Korea [1:57:38]:

Thank you, Mr. Chair. Before moving to box 2, I also would like to discuss box 1 briefly for the sake of our interactive discussion. And I would like to clarify my delegation's position on the issue of identification. In my previous statement, I acknowledged concerns about the possible exclusion of laws that involve human pre-setting of target profiles. To address this concern, I suggest specifying that identification occurs after activation, clarifying that it takes place while the law is being operated. But I also continue to hear the concerns that this approach would create a significant loophole by leaving our systems in which targets are preset by humans in advance. I'm not convinced by such argument. I would appreciate further clarification on that. And also, I share the views of the delegations emphasizing that our current work concerns definitions and characterizations, not regulations or prohibitions. While I agree with that in principle, it remains essential to lay out as concrete a foundation as possible. Without such a basis, we cannot build meaningful work upon it. And furthermore, I would like to join other delegations in supporting the proposal to add the words "they are integral" before "technological components" and to include the phrase "is designed to or intended to use" in order to better clarify the design aspect of laws. Now, on box 2, my delegation's comments will focus on paragraph 5 and specifically the term "context-appropriate human judgment and control". My delegation appreciates the Chair's effort to find a term that could attract consensus support. However, my delegation is of the view that the paragraph 5 is not the right place to introduce such a term. My delegation shares common understanding-- while delegations share common understanding on paragraph 1 through 4, paragraph 5 introduces a new concept that is both ambiguous and vague. and the phrase "context-appropriate" literally implies context dependence and therefore lacks concrete meaning. Introducing such a term as a legal requirement is impractical and difficult for implementation. We believe it would be impractical and even risky to place such a term on equal footing with other established rules and principles. We therefore suggest that this term be discussed in later boxes as one of many possible measures rather than in this foundational section. In previous discussions, some delegations argued that since we are merely considering possible elements and the boxes remain untitled, it is unnecessary to determine precisely where a particular paragraph should be placed. However, we must remember that, again, we are establishing the foundation for future negotiations and the foundation should be as logical and as structural as possible. Future negotiators will need to understand the context and how each element fits into the broader framework. In sum, my delegation suggests deleting the paragraph five at this stage and discusses the notions in later paragraph, later boxes. Thank you.


Chair [2:00:57]:

I thank the distinguished representative of the Republic of Korea for reacting to the remarks of other delegations on box one and for the views on box two. I would say whether we discuss it here or later, we have to discuss it at a moment, and this is clearly one of the main issues we have. If we just have human control, I think quite a number of people in this room are tempted again to come up with certain qualifiers or objectives. And I don't know whether that will be a meaningful discussion then. So this is why we came up with what we have now. And of course, we are open to other suggestions to see, to context appropriate, might not be precise, but we have, of course, also other notions that are not always precise, you could even say that they are in certain cases also context appropriate, like taking precautions, like proportionality. So let's see how the discussion develops on that. But I don't think we can do with just human control. I have two more requests for the floor at this moment, the United States and Australia. Let me first pass the floor to the United States. So you have the floor.


United States of America [2:02:29]:

Thank you, Chair. We're going to focus our intervention on two of the paragraphs within this box. We were very comfortable with the formulations that you have in box two, paragraphs one, three, and four. With respect to paragraph two, and this echoes a point that we've made in previous, both in formals and in a previous session, this is language that is framed on an ICJ advisory opinion, which itself uses a more specific framing of applicable law focusing on the cardinal principles rather than IHL more generally. And so we would prefer to see the opening clauses of this particular paragraph either focus on the cardinal principles of IHL or focus on the principles that we think are most relevant and applicable here, the principles and requirements of distinction, proportionality, and precautions in attack. So we don't have a specific line edit suggestion for you at this point, but wanted to register that view. With respect to paragraph five, we continue to have very significant concerns with references to context appropriate human control, meaningful human control. As we've noted in past statements, as we've noted in our working papers, we find the terms human control to be overly narrow. We don't think the terminology is appropriate for this instrument. IHL does not require human control as such. and software systems may be lawfully used to control autonomous weapons systems. We've grappled with this. This has been a constant theme in this forum since the very beginning. And we have found ways to try to address this in compromise formulations in the past. As the delegation of the Russian Federation notes, we have in the past found ways to refer to the control element. In the past, we have found ways to refer to judgment. And we also know that there is a reluctance to simply go back to language that we previously agreed. And so in the past, we've suggested focusing on paragraph 21 of the 2023 GG report, which does refer to control. and which we would be very happy to include here. That paragraph read, "Control with regard to weapon systems based on emerging technologies in the area of laws is needed to uphold compliance with international law, in particular IHL, including the principles and requirements of distinction, proportionality, and precautions in attack." I think all of us will remember that this was a paragraph that was very, very energetically discussed and debated and negotiated. And it represented a landing spot that was appropriate for a report. And I think we have a hard time understanding why it wouldn't be an equally important landing spot for elements of an instrument, which in our view are a more significant statement than language that's in a report. So we would continue to be very, very comfortable with that formulation. So, we have a structural concern with how the term is being used in the document. We have serious concerns with the approach of trying to adopt a new, vague, overarching standard, like meaningful or context-appropriate human control, and then to articulate a series of requirements that we think flow from that standard. We think this approach will inevitably create confusion and potential conflicts with existing standards in IHL that already apply. And we believe it's important first to articulate the existing IHL standards in this context, and then after that exercise, to consider what further standards, if any, could be adopted. So what language would work? We think, again, going back to the to the understanding that we've addressed control and we've addressed judgment in the past successfully. Unlike the delegation of the Russian Federation, we think there is a way to approach using both of those terms in a paragraph. And I think part of the key to that lies in their observation that this current paragraph does need to be broadened to include a recognition that the principles we're talking about apply throughout the life cycle. We actually think that could be helpful in finding a way forward. And our more specific idea at this stage in the discussion would be to frame the provision as follows. Measures, including during development, relating to control and human judgment are needed to ensure the use of laws is in compliance with the principles and requirements of distinction, proportionality, and precautions in attack, and other rules of international law applicable in armed conflicts. And so maybe to give a little bit more explanation of those ideas. First, deleting and effects. I think our view here is fairly simple, that compliance with IHL relates to human behavior, the use of laws in armed conflict, and not the effects of the laws. So that's our thought behind deleting and effects. The framing of principles of distinction, proportionality, and precautions in attack, and other rules of international law, including in armed conflict, is a formulation that's drawn from a number of instruments in IHL, including one that we're not a party to, Additional Protocol I. And so we think that framing at the end of the sentence is a more maybe a more rigorous way to approach the language of this text. And finally, framing this particular paragraph around measures that can be taken to control -- sorry, measures that can be taken during the development and use is a better framing and allows us to talk more specifically about control and judgment. Our delegation, I think our primary objection to the idea of setting out a meaningful human control standard or a context appropriate human control and judgment standard, especially in the way, Chair, that you framed it here, it appears in every single box. It is the core framing device for the entire instrument. It gives the impression that everything that's in the instrument flows from this. from this set of principles or for this particular principle. And as I've said, we just don't think that's true. But we've also said that we do see control as one of many measures that can achieve compliance with IHL. Our primary objective has been setting it out as the overarching prerequisite for compliance with the law. So framing this around measures that can include measures of control and measures of judgment allows for what we think is the fact that IHL does contemplate a number of different measures that can be pursued to achieve compliance. Thank you, Chair.


Chair [2:11:21]:

Let me thank the distinguished representative of the United States for sharing his views and going into depth, because initially I was a bit worried when you just referred to the discussion in 2023, because I also was present at that discussion. And of course, at a moment in time, then we run out of time, and we were clearly aiming at something to be agreed at that moment for in that report. At that moment, we didn't have an additional hour, and then we got three additional years. And I think we have really to make good use of the additional years. And I'm encouraged by your proposal where, in a different way, judgment and control are there. And there is also human. Because I think one thing, at least for me, is clear, that any text that comes out of this, somewhere there should be human in it, because this is essential part, of course, for the discussion we are having. with autonomous weapons systems, that there, the human role at a certain moment is not there, and that makes them so specific. So for that reason, we are dealing with it. Let me look at the list. We have -- and of course, please send them in writing, because we want to further study them and see what we can do with it, and also it would be useful for other delegations to have it and to digest it. I have Australia, the United Arab Emirates, Bangladesh, Belarus, but first Australia. Madam, you have the floor.


Australia [2:13:02]:

Thank you, Chair. We have been pleased to see that much of this box enjoys a level of support. From our perspective, we are supportive of paragraphs 1 to 4 as currently drafted. This includes the revisions to paragraph 4, which ensures that the Martin clause as reflected in the rolling text, aligns with the preamble of the CCW and clarifies that international agreements such as the Geneva Conventions do apply to laws. Chair, Australia proposes the deletion of paragraph 5 from this box for the following reasons. First, the mention of context-appropriate human control and judgment belongs in box four, which sets out measures specific for laws. The focus of box two is on the principles of IHL, while box three elucidates obligations and prohibitions of IHL as they apply to laws. Context-appropriate human control and judgment is not a principle of IHL. nor a prohibition or obligation. Rather, it is a measure to support compliance with IHL in the use of laws. Second, paragraph 5 introduces confusion, given it is similar but also different to box 5-- sorry, to box 3, paragraph 6, which also discusses context-appropriate human control and judgment. Deletion of paragraph five at this box would avoid two different formulations in two different boxes, which then raises the question of why there is a discrepancy. Finally, irrespective of placement, Australia would prefer to return to the original formulation of context-appropriate human control and judgement. We consider that context-appropriate should be more closely linked to control as compared to judgement, which is always needed. Thank you.


Chair [2:15:19]:

Let me thank the distinguished representative of Australia for sharing positions and indeed also already to reflect on the boxes three and four because that would have consequences. Next on my list is the United Arab Emirates. Madam, you have the floor.


United Arab Emirates [2:15:42]:

Thank you, Mr. Chair. We appreciate the interventions of the countries which in their majority concur with the position of the UAE. Regarding paragraph one and two, we welcome the mentioning that We need to respect Article 55 of the First Protocol of the Geneva Conventions. We also concur that the international humanitarian law is applicable to all conflicts, all weapons, present weapons and weapons in the future, regardless of the technological component. And this is an essential pillar to protect civilians. Regarding paragraph 3, we support the formulation which states that the international humanitarian law applies fully to the development and use of laws. And we reiterate the importance of not overcoming this context under technological pretexts. Regarding paragraph 4, we appreciate mentioning the principles of humanity and IHL according to Martin's clause, and we recommend to have a formulation that reiterate that legal commitments should not be exempted in case of absence of law because we want to guarantee that there would be no legal void or vacuum. Regarding paragraph 5, this paragraph meets the position of my country, which is context appropriate human judgment and control in line with IHL. However, we may need a clearer formulation regarding the timeframe for this obligation, be it prior to the use, during the use, or after the use of these weapons, as well as defining the functions of control, such as accountability of behavior and others. We also would like to mention that there is no reference to a legal revision prior to the use of laws as per Article 36 of the first optional protocol of the Geneva Conventions. And this should be applied prior to the use of the weapons. So this should also be linked to the principles of accountability and full compliance and strict compliance with the international humanitarian law. Thank you.


Chair [2:18:56]:

Let me thank the distinguished representative of the United Arab Emirates. And next on my list is Bangladesh to be followed by Belarus and France. But let me first pass the floor to the distinguished representative of Bangladesh. You have the floor, sir.


Bangladesh [2:19:15]:

Mr. Chair, thank you for allowing me to speak. As my delegation takes the floor first time, I deeply appreciate your bold leadership and promise our all-out support to you. It is good to see that the rolling text box 2 has stressed on legal principles invoked for regulation of laws. It has duly emphasized on the compliance of IHL. The rolling text has also mentioned individual accountability in box 5. However, it has so far stopped short of expressly referring to individual criminal responsibility, one of the core principles of international criminal law, ICL. I am detailing on why my delegation considers it important to mention this principle in box 2. Firstly, As we know, this principle has addressed a loophole in international law. It ensures that individuals are criminally responsible under international law for international crimes, including war crimes. It acts as a strong deterrent to the unlawful methods and means of warfare by holding the individuals criminally responsible for their crimes. For the effective protection of individuals during armed conflicts, coherent and complementary application of IHL and ICL is essential. Secondly, laws by its very autonomous nature stands to remove the humans from its operation, which may even involve the kill decision. Decreasing or removing the context-appropriate human judgment and control will make determination of individual criminal responsibility difficult in the event of serious violations of IHL by laws. It will open legal loopholes for evading individual criminal responsibility for war crimes and encourage states to desperately proliferate, deploy, and use laws during armed conflicts. Therefore, unregulated laws stand to directly damage the very basis of international criminal law. My delegation believes that renewing our commitment to the principle of individual criminal responsibility will strengthen our efforts to ensure human involvement in the operation of laws, Mr. Chair. Finally, Strengthening the international legal order is a never-ending collective process. We have a sacred duty in this regard, especially while framing new arms control instruments. In this case, we can contribute by expressly referring to the principle of individual criminal responsibility, particularly in box two. The suggested text could be as follows. Individuals, not machines. bear individual criminal responsibility for serious violations of IHL. The text could be further enhanced by the distinguished delegates. I thank you, Mr. Chair.


Chair [2:22:23]:

I thank the distinguished representative of Bangladesh for stating the views of his delegation and for making proposals to the meeting. Next on my list is Belarus, to be followed by France and Austria. But let me first pass the floor to Belarus. So you have the floor.


Belarus [2:22:42]:

Thank you, distinguished Chair. As it's the first time that my delegation has taken the floor, I would like to add my voice to those of other colleagues who have expressed gratitude to you and your team for the skillful leadership of our group. like to congratulate you on your high degree of professionalism and your desire to seek compromise and prepared to work on the working paper. And we'd like to thank you for all of the many useful consultations. The comments from the Belarus delegation concern para 5 of box 2. The issue of control and responsibility for the use of laws in violation of IHL is key in our discussions and is extremely complex. International humanitarian law violations might be the result not only of something which has happened during a war, but also in the design of the system and the decision to activate the system. And so responsibility might not only lie with the combatants who directly used laws to attack the opponents and the commanders and those who took the decision to activate laws, but it could also be the engineers who worked on developing the laws, the programmers who developed specific algorithms, the consultants who worked on laws and technical staff who serviced the system. Also, we should bear in mind the fact that in developing laws, we must make sure that it can be operational, and that might not be a sufficient guarantee that it is in compliance with international humanitarian law. Given that, we feel that it would be right to define the concept of control through the factual use of any measures to make sure that IHL is used in compliance with the norms and standards of international humanitarian law. If IHL is not in line with these standards, there could be bad faith competition between countries that develop and use laws, and parties might try to introduce legal prohibitions and to avoid limitations which might create difficulties for them. It is important to leave complete freedom to states to use laws with any technology, methods and approaches, so long as that is in full compliance with international humanitarian law for such systems, and full legal responsibility for such states for any of the effects of their laws. Given that, we suggest that para. 5 of box 2 should be drafted as follows: 5.1 Control is ensuring the use and/or functioning of laws in compliance with the principles and standards of international humanitarian law. 5.2 The use of laws without control is prohibited. Paragraph 5.3. If laws is used in violation of IHL, the country where the laws in violation is located shall carry out the corresponding investigation to identify violation of IHL by individuals and to take all measures necessary to ensure there's no repeat violations of IHL in the use of this laws. Full stop. Thank you.


Chair [2:26:48]:

I thank the distinguished representative of Belarus for stating the position of his delegation and making proposals. I think some of them are also a bit already pertaining to box number five on responsibility and accountability. Next on my list is France, to be followed by Austria, Pakistan, United Kingdom, and then I have still campaign to stop killer robots and the Philippines. But let's see how far we come. First, I have France, who's available.


France [2:27:23]:

Thank you, Chair. Box 2 generally suits us, and we support these five paragraphs as they stand. Turning to the discussions on the expression "context-appropriate human judgment and control", at the national level, France uses the term "human control", which covers all these different forms of control. But we can be flexible as to the terminology and support your compromise expression "context-appropriate human judgment and control". We have explained in a joint non-paper presented by Germany this morning what our understanding is of this expression. IHL does not explicitly contain rules guaranteeing human judgment and control. The use of weapon systems always requires some form of human intervention, which can be indirect control, in order to guarantee compliance with the different rules and principles of IHL. In this paper, we describe firstly the different elements of the expression, and then we explain why we believe it reflects very well the concept of human involvement. Then we give details as to how this concept can take different forms at different stages in the life cycle of different weapons, including the development stage. Finally, we talk about the different context aspects which need to be taken account of so that the appropriate level of human judgment and control hinge to a great extent upon these different context related factors. Thank you.


Chair [2:29:14]:

I thank the distinguished representative of France for stating the position of our delegation and referring to the paper earlier. Next on my list is Austria, to be followed by Pakistan and United Kingdom, but first the distinguished representative of Austria, you have the floor.


Austria [2:29:36]:

Thank you very much. Um, from our side, we could also, um, support the current formulation of COX-2 as it stands, maybe with one, uh, caveat that we have some question mark on the change or the reversal of control and judgment. In the previous version, we used context-appropriate human control and judgment. This has been changed in the meantime, since this is the first time that we are discussing this in a plenary and an official format. We would just like to hear, for those delegations who are in favor of this change, what this actually adds. In the end, the way that we understand it is that this refers to human control and human judgment. or human judgment and human control. If there is a notion that the human would not relate to control, we would have a problem with this reformulation and would rather prefer the previous version. I think the control is -- the adjective human is much more important when it comes to the control than to the judgment, because in the judgment, the human element is inherent, while with the control, I think it's much more important to spell this out. So the proponents of this change I would ask them to argue why this is the better formulation than the previous one. If there are no arguments for it, I think we can also go back since the other one is resolving more problems than this one. Just then reacting, since this is our only point, reacting to some points in the discussion, I think the proposal from Bangladesh to also have addition of international criminal law is interesting. In general, we see box 2 as setting out either elements that could be part of a possible preambula or elements that relate to international law in general, so it might be the right place. But at the same time, we are also aware that there is some reference to international criminal law, I think, in box 5. On the other hand, again, we have a lot of, uh, agreed language when it comes to accountability and responsibility, there's a rich body of work on this topic from previous iterations of the G20 on laws. I think since we are having this, this kind of a key, it wouldn't damage to have also something on this in, uh, box two, since this is kind of the, the, the, the higher, uh, level principles that we are referring to. Then other reacting, since Korea asked for specific examples of loopholes, I think we provided some of them already. Going back to box one, since it also came up again, to explain again, our biggest concern is since identify is a rather broad formulation and that does not specify at which point in time and where it happens. and also not gives more clarity to the interpretation of this word. It could mean that we are taking specific weapon systems out of our definition simply because the identification aspect is not part of its operation. It could happen in general because of its design, but it could also happen that, as I think Article 36 explained, that the same system in some use has the identify element in it and in other uses not have it. So we are creating a system that moves in and out of our definition in different ways of operating. And I think this is probably a general problem that we are facing, and once again, an argument for a broader interpretation of our definition and characterization. Coming to other points that have made when it comes to para 5, in our interpretation, and we've made this argument before, para 5 is linking the general principles of IHL with many elements that are following. If you argue that the box two will be a preamble, you could actually shift it to the beginning of box three. But in general, we still think that this is the better place since it's creating a link of the principles of IHL that we are setting out before with the key concept of our work, which is either the human control or, in this case, the context appropriate human judgment and control. And as you have also pointed out, or other colleagues have pointed out, this is a continuation or a further development of a concept that was an outcome of the 2023 work. It was actually the most important, and some might argue even the only essential outcome of our work during the 2023 mandate. So it is key that we are looking further into it and are developing it further. And I also fully agree with your assessment that the reason why it is a rather rather lacking in substance, Bara, is not that this is what we wanted to have as a final result, but just because we run out of time at the end of our discussion. We are also not fine with just using the language of 21C of the 2022 free report, it is important that we move forward and that we actually add some flesh and, and give it more, um, substance and explain what we mean and how it is, um, to be used. When it comes to the general formulation, I think if we're going back to our past positions, we would have agreed with the Russian position and would have argued that we should focus on human control or on meaningful human control and then elaborate how meaningful human control can be achieved. And this was also one of the main content of two of our working papers. I think this is still a worthwhile exercise, but you also know that we have moved forward with our discussions, and to some extent this is still there. We, and as I said this morning already, we appreciate that many delegations could live with the formulation of context-appropriate human control and judgment, and feel that our discussions have progressed in this regard, and we have finally a formulation where many delegations from different original positionings can rally around. So we welcome this development. We also feel that it resolves some problems, especially when it comes, for example, to defensive systems, but also when it comes to systems that are used in an environment that is rather not endangering the civilian population or creating other problems. But as some delegations have pointed out, it also creates new problems, especially when it comes to its specificity and unclarity, what it actually means. But we see potential in discussing this further. We also have seen a working paper provided for this purpose by some countries, and I think this is extremely useful. In general, we believe to give more clarity on what we mean with context appropriate, we should give more examples and also clarification on what factors are relevant in this context. We were also working on some, but we could not finish it in time, but maybe we will come back to this in the future. Another proposal for the consideration of other high contracting parties that we made is that those examples on what context appropriate human control and judgment mean could be added into a technical annex. I think this is also in line with previous examples of how our convention works. And it could give some indication on, and more clarity, since this is, uh, actually the, the biggest problems that many, many allegations see, and, um, to some extent, um, rightfully. Finally, and this is also a point I wanted to make, we are a little bit surprised that some delegations have no problem with giving a definition of lethal and what lethal means, which has absolutely no root in IHL, but at the same time have concerns over the term of context appropriate. I think there is a little bit of a logical problem here. Lethality is a highly problematic definition to make, while context appropriate gives us some leeway and gives us some opportunities. I think there's some inconsistency here when it comes to the positions of some delegations, I have to say. Thank you.


Chair [2:38:11]:

Let me thank the distinguished representative of Austria for stating the position of the delegation and also engaging on previous remarks. Next on my list is Pakistan, to be followed by the UK and the campaign to stop killer robots. But let me first pass the floor to the distinguished representative of Pakistan. You have the floor, sir.


Pakistan [2:38:34]:

Thank you, Chair. It's very easy for me to follow Austria, since I can say, of course, that we are very much in line with the arguments that they have made on this box. I would just like to make some additional comments and specifically on paragraph 5 of box 2, uh, first of all, to the questions, uh, whether we should debate about the placement of this concept here in this box or not. We believe that perhaps a discussion on what is the right place for this concept right now may not be conducive, because for us, we do think that the boxes have been organized and presented in a way to facilitate discussions and not to, of course, prejudge where certain elements would go in the instrument, in a future instrument. And we have, of course, our preferences. We would like to see in a future instrument where these important concepts would go, and we would come back to it. But as of now, we are just addressing the substance of it. So in this context, let me address the question first, as some have pointed today and in the past also, that the phrase "human control" is not explicitly present in the IHL text, therefore we should not consider it here. We have argued in the past that the absence of explicit language in historical treaties, international law, or IHL do not speak of its irrelevance today. Rather, it highlights the unprecedented nature of the questions that have been posed by autonomous technologies. Until now, human control has always been inherently assumed in the weapons of past. It was intrinsic to all such weapon systems. So if we had found refuge in this argument in the past that certain terms do not exist in international law, we would never have found the need to progressively develop international law or IHL in the first place. Now, Chair, let me address the question whether we need both terms, human judgment and human control, or whether we can perhaps substitute one over the other, as has been suggested today. In our view, these two terms, they do different jobs. Human judgment, we think that it's cognitive, legal, ethical assessment required by the IHL, for instance, on distinction, proportionality, and feasible precautions. It's about a human deciding whether a particular use of force can be lawful and acceptable. When we think about human control, it's the practical capacity and arrangements. And for instance, the process of setting constraints, limits. So judgment, in other words, is a decision and control is the means. And you need both to have to meet the IHL in practice. And when we look at the history of development of these terms, in 2019, we agreed very clearly that the human judgment is essential in order to ensure that potential use of laws is in compliance with IHL. And in 2023, as has been referred by certain delegations, we reached an important milestone where we recognized control as well. So to the question that why we cannot perhaps just restate what we agreed in 21C of the 2023 report. For my delegation, I would also like to recall that we agreed at that compromise, of course, when we were running out of time to agree on a report, but we cannot look at that in isolation. It needs to be seen in the context of all past agreements that we have done in this area. That includes the 2019 report. That includes the future instrument, of course, all the elements that we have already agreed on, for instance, the guiding principles. So it's important that we build on the previous agreements and not just restate our past convergences. Now, uh, let me also briefly address the term context appropriate, uh, change. Chair, you would recall that when this debate started, my delegation was one of those States which had called for retaining the word "meaningful human control". However, we understood that that was termed as perhaps vague or potentially very narrow by certain delegates. Therefore, to address those concerns, this context appropriate was introduced in the first place. So the term context appropriate is meant to ensure flexibility and acknowledge that the extent and form of human involvement necessary for compliance with IHL will naturally vary based on operational context, technology, and scenario-specific demands. In the past, we have talked about certain, those scenarios and environments, what the context would be, for instance, the operational environment. For instance, the complexity and clarity of targeting decisions. For instance, the nature and duration of the missions. We are open to listing a non-exhaustive list of these context or operational environment in a technical annex in the future. But I think it's important that we give that flexibility to implement and exercise human judgment and control. And lastly, I think it's important because it connects also with our previous agreement in the guiding principle, see that when we had said that in determining the quality and extent of human machine interaction, a range of factors should be considered, including the operational context. So this context appropriate is an attempt to define and further implement that guiding principle where we are talking about what kind of scenarios could exist. In short, in a nutshell, we think that this could be this phrase context appropriate. Human judgment and control represents the best available compromise given the divergences and positions that have been expressed in the past. Thank you.


Chair [2:44:53]:

Let me thank the distinguished representative of Pakistan for his intervention and also for making a link to the text and some of the proposals made in box 3 under paragraph 6, because that is clearly there. I'm having a look at my watch. I think we should stop here for today. Then tomorrow morning, we will start with the United Kingdom. Uh, followed by the campaign to stop killer robots, the Philippines, Italy, Brazil, and Israel, and we have one more, Colombia, so already quite a list, um, I think there is also some food for thought for you, um, with regard to, um… proposals made today with regard to the text, I would encourage delegations to circulate their proposals in writing because some of them were quite complex and we really, I think you all deserve some time to, you all deserve some time to consider them carefully. Clearly, we are now a bit in a school room phase, everybody's already packing up. That would almost bring me back to university, where we would say, "Nunc est bibendum." Indeed, the reception is ready as of 6 o'clock. So let's slowly, slowly move. It's just adjacent to this room. I would like to thank you all for your engagement, for your constructiveness, for your proposals. for also the interactive debate. We will continue tomorrow. I would also like to encourage you to make good use of the reception, to indeed build on consensus, talk to each other, try to understand each other, and then let's continue tomorrow. And also my sincere thanks to the interpreters for their work this afternoon. See you tomorrow morning. This meeting stands adjourned.