4th Meeting - 1st Session Group of Governmental Experts on Lethal Autonomous Weapons Systems 2026

Excellencies, distinguished delegates, I ask you to take your seats again. Okay, let's start. I have a number of requests for the floor. I have Australia, Argentina, the Russian Federation, France, Canada, Israel, Norway, Utah University, ICRC, and somewhere Uruguay, depending on-- oh, and we have one more. Pakistan? Yes, okay. But let me now start with giving the floor to the distinguished representative of Australia. You have the floor, Madam.

Thank you, Chair. Australia will deliver two interventions on box three. In our first intervention, we will focus on paragraphs one to five, as well as paragraph eight. In our second intervention shortly, we will focus on paragraph 6 and 7, which concern context-appropriate human judgment and control. On box 3, paragraph 3, we would propose a minor edit to resolve the issue that civilians directly participating in hostilities remain civilians, but may be targeted under IHL in defined and specific circumstances. We suggest adding the words "protected as such" after "individuals." On box 3, paragraph 4, we support the inclusion of the second sentence, as we do see practical utility in setting out the limited circumstances in which a state could interact with laws that cannot be used in compliance with IHL, for training to identify, disarm, and destroy these laws, and for the development of countermeasures. We have heard some delegations raise questions as to how this would be implemented in practice, and are open to textual edits to make this implementable. Further, Chair, as the final elements, however presented, would no longer contain the boxes, we would like to offer the following suggestion to provide clarity when the boxes are read together. We propose that the key components of box 3, paragraph 8, be combined with box 2, paragraph 4. As such, box 2, paragraph 4, would read: IHL applies fully to laws and the requirements of IHL must be taken into account in the design, development and use of laws. As a result, we could then remove paragraph 8 from this box. By being clear that the requirements of IHL must be taken into account in these three stages, we ensure the requirements of IHL are appropriately strong and addressed upfront. This provides greater clarity and avoids unnecessary duplication. I thank you.

Let me thank the representative of Australia for sharing the views of her delegation, making proposals also with regard to the combination of powers. Next on my list is Argentina. to be followed by the Russian Federation. But let me first pass the floor to the Ambassador of Argentina.

You have the floor, sir.

As we stated in our previous statement, my delegation considers that the text is at an advanced stage of development, with certain elements still pending, and they are fundamental Although there are still divergences, and in this section is certainly one of those. Now, over this process, we've proposed specific language on the human intervention. It is important that section 3 maintains reference to human intervention. We are flexible on the final wording on the level of control of human beings in the process and on its specific position, so we are open to considering that these provisions may be applied, as is the section on IHL, which was discussed last week by the US delegation. There's also a wording of paragraph four, which is more direct, referring to the manufacture, acquisition, stockpiling, and transference that cannot be used in compliance with IHL, where the rest of the paragraph would be deleted. referring, deleting the reference to military applications. That could be complicated for my delegation. I thank you.

I thank the Ambassador of Argentina for stating the positions of his delegation. Next on my list is the Russian Federation, to be followed by France and Canada. But let me first pass the floor to the distinguished representative of the Russian Federation. Sir, you have the floor.

Thank you, Chairperson. We also have a number of comments to make on section 3. First, Chairperson, we continue to insist that in stylistic terms, 1, 2 and 3, sub-paragraphs 1, 2 and 3, need to be reworded so as to bring them more into line with the mandate of the group. And that's appropriate, so stylistics be applied that are also in line with the report of the group. They should not look as legally binding provisions. with prohibitions and therefore they should not be starting with the words "it is prohibited". As we mentioned before, the same meaning could be conveyed in quite a different stylistic way in these subparagraphs. If, for example, we take paragraph 1, then instead of saying it is prohibited in all circumstances to use, at the end we could add the words "must not be used", and that would be sufficient. On para subpara 2, instead of "it is prohibited to use", we could also add the words "must not be used" at the end, and in 3, Instead of "it is prohibited to make", in the middle of the subparagraph, we could write "should not be the object of attack by laws". Secondly, Chairperson, under subparagraph 2, we to categorically do not agree with the comments by a number of delegations, namely that the foreseeable consequences of the use of laws effects are a provision or a concept established in IHL. That is not the case. And as we see it, talking about prognosis or prediction of the effects would not be relevant in the context of IHL, in the context of subparagraph 2. So we propose deleting "anticipated" and leave just the verb "limited". Furthermore, in this subparagraph, we continue to insist, and that has been the case over many sessions of this GGE, as regards adding the wording which we all adopted in 2019 in the report, Under paragraph 23, on preventing risks at the stages of design, development, testing and deployment of lethal autonomous weapons systems, we propose taking the wording from paragraph 23 of the 2019 report in full. both sentences, and in our view, they are entirely relevant. Furthermore, they are very much needed, given the comments that from session to session we hear from many delegations. So this also applies as regards the phases of the life cycle of these systems. in design, development, testing and deployment, and not only in use of these systems in armed conflict. Once more, I emphasise that we again propose and insist on the inclusion of the agreed language in paragraph 23 of the 2019 report into sub-paragraph 2 here. On sub-paragraph 4, we align with the comments and concerns expressed by many delegations today as regards this sub-paragraph. As regards the second sentence, we, like many other delegations, do not understand. the first sentence in the subparagraph. We find it to be superfluous and indeed counterproductive, since we have section 4, paragraph 1, where we have a clear definition of obligations in terms of legal review covering the stage of design development and other subsequent stages, since, in the context of obligations on legal review, no weapons systems which are not in compliance with the principles and standards of IHL can appear. They cannot be developed, they cannot be approved, and they cannot be transferred to anybody. And so, paragraph 4, sub-paragraph 4, it gives a distorted view of reality where this situation could appear to be impossible, even in theory. That is not the case. And thus we are categorically opposed to sub paragraph four and propose deleting it in full in line with what has been proposed by a number of delegations before us. Now on sub paragraph six. First of all, we'd like to look at the topic of the context of the human judgment and control, context appropriate human judgment and control. From session to session, we discuss this concept and still haven't reached any unified understanding. And as we see it here, we should be talking about two sets of issues. The first, as regards the wording context appropriate, because this wording, again, is not clear. It's still not clear. And we believe that it has been put forward by a very limited group of delegations. It is not shared by the majority of delegations. So we propose abandoning that concept. We are prepared to follow the proposal of India and a number of other delegations who suggested opening up this concept in order to determine what is behind those two words. Our delegation is prepared to consider any proposals in that connection without in any way prejudicing the final result of the discussion. And in previous years, in the reports of the group, it was stated several times that a certain context and a certain set of conditions for the holding of military operations can have an impact on ensuring human control. Now, the second set of issues is connected with the context of human judgment and control. Human control, human judgment. Our delegation is prepared to be flexible on this. We like the concept of human control as a separate concept. and likewise the concept of human judgment. And we're prepared to agree to bring those two concepts together in the wording "human judgment and control". But what's clear is that the human element must be clearly set out, since for many delegations, if not for all of them, This concept is a central one. And on paragraph 6 again, we would propose somewhat amending the following wording. We believe it is essential to were the beginning of the paragraph to make it clear that we're talking about with human judgment and control, that it covers all lethal autonomous weapons systems, and not only the issue of guaranteeing compliance with international law, in particular IHL. as regards application and concept and the effects of the use of such systems. And so, in this subparagraph, we could say the following:.

"Human judgment and control with regard to lethal autonomous weapons systems is needed to ensure compliance with international law, in particular international humanitarian law, including the principles and requirements of distinction, proportionality, and precautions in attack.

Coming to subparagraph 7, here again, if we look at the chapeau, the same problem is there as regards context-appropriate human judgment and control. On sub-paragraph, although, Chairperson, if we look at the whole sub-paragraph as a whole, then on the specific parameters of ensuring human control, which is what we are talking about here, our position has not changed. we believe that these parameters for ensuring human control remain at the discretion of the State Party to the Convention. And thus, if we take the specific sub-paragraphs here, then we need to follow a very fragile balance between providing flexibility and freedom to States Parties in the wording of the parameters and then establishing the parameters to ensure human control and managing the specific concerns from a number of other delegations on establishing specific concepts as to how, in one or other set of circumstances, this control can be ensured. And moreover, on the specific wording, there also needs to be an avoidance of duplication from previous or following paragraphs in this document, in substantive terms above all, but also We need to prevent any excessive wordiness or stylistic overload in paragraph 7 with its subparagraphs in order to ensure that they are clear, that they are precise and specific and can be understood not only by the States parties, but also by observers and delegations that are not party to the work on this Convention. And looking specifically at the subparagraphs now.

So, paragraph a.

Our delegation is ready to agree with the first sentence of that subparagraph. As to the second sentence, here we think on the one hand It lacks a sort of tangible nature. There's something tangible missing. What's more, it contains excessive substantiation of how human control could be carried out. In our opinion, that is unnecessary. given that in section 2 we clearly state that IHL is fully applicable to the use and development of laws and it's entirely sufficient to limit paragraph 8 to just the first sentence. That's how principles and norms of IHL will be implemented, the specific chain of command and control should stay at the discretion of specific states and of specific armed forces, depending on the national realities that a given state party to the convention has. Therefore, we propose deleting all of the second sentence. It is completely unnecessary and only muddies the waters. It does not provide any clarity. What is more, the concepts such as ethical considerations, let me remind you that the one time that that concept is used in the report of the group And that its reference is not, it's nothing to do with the use of or development of laws, rather it refers to work in the context of this group. That's the only time ethical considerations are referenced. The work in the group, in the context of the work of this group. On turning to B, pass on to B. We think it looks currently Like a bit of a Christmas tree. We see there's been an attempt to include in this paragraph to include many of the proposals that were put forward by delegations. I don't think we need that. The first part of that subpara to us. should clearly define the limitations with respect to which this sub-para could pertain to. In our opinion, those limitations should cover the scale of operation, type of targets, operation duration, and geographical scope of the operation. That is the substance and the core of the paragraph. We need to highlight and clearly say that we need to limit the scale of the operation, the types of targets, the duration and the geographical scope. Everything else there is completely redundant. Then on proposals regarding the number of engagements or use of those kinds of weapons, we still don't understand how that will allow us to ensure due human control. We do not understand how that will allow us to prevent or mitigate risks when it comes to use of such weapons during hostilities, so we propose deleting the tail end of that paragraph. Accordingly, as we said, this sub-paragraph should read as follows.

Limit the scale of the operation, the types of targets, duration, geographical scope.

Turning to C. We agree with many of the comments made by many delegations before us. This sub-para, on the one hand, it just doesn't look like it's been worded properly. On the other hand, also, That sub-paragraph is inconsistent with the realities of this day and age that we have on the ground. In that context, machine learning is already being actively used in specific weapon systems, including machine systems that are conducted in real time. and we think that including that sub-paragraph would run counter to the understandings that we reached earlier in this group with respect to the fact that it was counterproductive to limit technological progress. We believe that that sub-paragraph is specifically aimed at restricting technological progress. Moreover, if machine learning If amending mission parameters by specific weapons systems with a high degree of automation or fully automated weapons systems is to be properly conducted, taking into account and in line with the principles of international humanitarian law, we can hardly be saying now that that will not be consistent with the existing norms of international humanitarian law. We propose completely deleting that sub-paragraph. On sub-paragraph, our position has not changed. We do not think that it is necessary to go into specific details about the deactivation of such systems And therefore, we think it's possible to stick to just the first part of that sub paragraph up to the word in a timely manner. And delete the second-half of that paragraph, as well as give individual cases of how such deactivation could take place. There's many ways. to deactivate a system is not just with the help of self-deactivation, self-neutralisation or self-destruct mechanisms. Those are specific cases, and going into their detail is unnecessary and only creates added confusion. On E, we share many of the remarks that have been made before us. We think that that sub-paragraph is repeating paragraph 7b, as well as paragraph 3 in the same section. We do not think that it provides any added value and that it should be deleted. Paragraph 8 is redundant. Many other colleagues have said that. We have paragraph 7, which already fully addresses all phases of the life cycle of those weapons systems. If you look at the chapeau of that paragraph, it does not say that it is devoted to any specific phases in the life cycle. We think that paragraph 7 is entirely sufficient. It applies to all phases of the life cycle in the context of appropriate human control. Paragraph 8 does nothing to add to paragraph 7 and should be deleted. A number of delegations have commented on restoring in paragraph 7 or in section 3 as a whole the concept of predictability, reliability explicability and traceability of those weapons systems. Our position is as it was before. I don't think it would be appropriate at all to reintroduce that concept in this section. That concept has nothing to do with international law, nor will it have anything to do with international law, we think, in the context of that category of weaponry. That concept is not clear to us. If any attempts are made to factor in that concept, we will have to establish what exactly stands behind those mysterious words, which are traceability, predictability and explainability. We will have to establish what is behind them, because we do not yet understand what stands behind them. We have heard no clear explanation of what is meant by those terms. In conclusion, we would like to react to some of the comments made by the US delegation. We will not touch on specific proposals that have been made; we just want to share some of the concerns and misgivings that our delegation has in the context of what we heard from the US delegation. especially as those misgivings are getting greater. The fact that we are only hearing about those issues at this late stage, when we are discussing specific understandings that could lie as the basis of the report for the group, is worrying, especially in a year when we have the review conference, which is there to review the results of discussions within the group. and our misgivings are connected chiefly to two things that we've heard today from the US delegation. The first thing was putting under, in question the whole concept of human control, calling into question the whole concept of human control. We do hope that the comments that we heard have the intention of completely erasing all the work that has been carried out in the group with regard to human control, and that they will not call into question the fact that that concept is one of the central concepts in the context of the work and discussions in the group. If that is the case, I fear that we sliding way back. We are regressing a long way and, effectively, we are basically abandoning the work that has been done by the group. If that is the case, we do not know how we should conduct ourselves in future. Our second misgiving is related to the comment about previous understandings of the reports and considerations that had been agreed within the GGE. We also hope that the US delegation is not stepping back from those previous understandings that were reached in the group. that all group reports that were agreed by consensus, especially the guiding principles that were approved by the states parties, remain in force. We trust that they still remain relevant and are fully shared by the US delegation. If that is not the case, we would like to hear that now, because that would be a serious change to the work of the group. And we think we would then need to discuss entirely things entirely different to the document presented by you, Chairman, because this would call into question, it would jeopardize implementation of the whole mandate of our group. Otherwise, thank you.

I thank the distinguished representative of the Russian Federation for sharing the views of his delegation, for making concrete proposals with regard to the text with regard to the rolling text as well as reacting to the statements of other delegations. The list has been growing substantially. Next on my list is France, to be followed by Canada, Israel, Norway and Utrecht University. But let me first pass the floor to the distinguished representative of France. Madame, you have the floor.

Thank you, Chair. On section 3, we would like to recall some of our positions and react to the proposals made during informals and this morning by other delegations. Paragraph 4, if we could have it on the screen, please. Like other delegations, like Sweden, we are struggling to understand the scope of this measure. It seems important to us to ensure at least that it would be possible to develop certain systems, not with the intention of using them, but so as to develop countermeasures. Therefore, so as not to overburden research, it might be useful to integrate this issue of the intention of use in the first sentence of the paragraph. On paragraph six, the notion of human control is central to our debates and to the coordination of a future instrument. It's true that no treaty rule applicable to the conduct of hostilities mentions expressly the notion of human control. However, we believe that it derives from those rules that the implementation of IHL depends on a human ensuring the conformity of his actions under the use of force and ensuring that they are consistent with the rules governing the conduct of hostilities. Paragraph 6 includes the crucial notion and measures for application are then listed in paragraph 7. Paragraph 6 we can live with as is, but the solution proposed by Australia could also be a good idea to resolve differences in the room. However, we think that a number of subparagraphs in paragraph 7 integrate elements that seem to go beyond our obligation under IHL. Those elements should not be included in the text, as they are currently. We would also like to keep the conditionality of the measures that are listed in paragraph 7 in the context in which they would be applied, and therefore to keep the term "should" as the delegation of Spain suggested before. In sub-paragraph, HIL does not call for reasonable vigilance when it comes to ethical considerations. It calls for a commander to evaluate a situation and take a decision on use and parameters of use based on legal regulations and operational considerations, command can decide on a case-by-case basis to consider ethical perspectives. Therefore, as Italy, Singapore and Colombia said, we could support the proposal of Japan here. Subparagraph. If IHL calls on the commander to anticipate the impact of a weapon, it does not, however, call on him to limit the number of engagements. That limitation can be part of the usage parameters on a case-by-case basis, as defined by the commander. Moreover, the text currently creates certain confusion with the scale of the operation. which here seems to cover the type of targets, duration and geographical scope, whereas in other paragraphs it is different. Therefore, we could support the proposal of Japan that would resolve that problem. Paragraph C The mission parameters of laws should be defined in advance by the commander, and then the machine learning inherent to certain systems could establish certain parameters without, however, calling into question the framework of operation or the critical functions of the system. Therefore, it is crucial for the commander to be able to reassess modifications when they are likely to modify his initial appreciation or understanding of a given situation. If machine learning does not go into question that system, then human intervention would not necessarily be required. Therefore, it seems important for us to keep the qualifier here significantly. or to replace it maybe with another such as substantially, which was raised during informals, suggested during informals. It's also important to reintroduce with regard to the types of targets, duration, geographical scope and scale of operation as proposed by Singapore. Subparagraph in E. As I have already said, the parameters must be defined beforehand by the commander. It is he who will define the mission parameters. He should take into account duration, geographical scope and type of target, and not just a geographical perimeter. It must restrain the target to military targets. in terms of location, use and targeting. Therefore, it would be good to reintroduce the clarification with regard to the types of targets, duration and geographical scope of the operation. Moreover, it is important not to interpret that paragraph as limiting the use of systems for only military objects by their nature. Laws can be used in line with IHL against military targets, but also against combatants, armed groups or persons who participate directly in hostilities. In that respect, we could support the proposal made by Singapore and Japan to replace military objects with military objectives. Finally, paragraph 8. We think that restrictions should be taken into account during the development phase. However, I don't think it's It is not necessary to take into account all the measures listed in paragraph 7 during the development phase. Obligations and voluntary measures, which are introduced in paragraph 7 by the word "should", should not be placed on the same footing. We heard the proposal from Australia, and that seems relevant in resolving the problem that we have with this paragraph. I thank you.

Let me thank the distinguished representative of France for making proposals and commenting on other text and text proposals. Next on my list is Canada, to be followed by Israel, Norway, and Utrecht University. But let me first pass the floor to the distinguished representative of Canada. You have the floor.

Thank you, Mr. Chair. My delegation wishes to make some remarks on the paragraphs found in box 3. Firstly, in paragraph 2, We support the replacement of "controlled" with the term "limited," which brings us more in line with the IHL principles of proportionality and precaution. In paragraph 4, we have listened carefully to delegations' concerns, particularly with respect to linking the acquisition, stockpiling, and transfer of non-compliant laws to international humanitarian law. We agree that these concepts are not themselves found in IHL. Our delegation understands that systems that cannot be used in compliance with IHL would be unlawful to employ in armed conflict and that it follows that there should be no reason to manufacture, acquire, stockpile or transfer such weapons. However, we also recognize that limited retention for the purposes such as training and the development of countermeasures is an important and legitimate measure for responsible states. For these reasons, Canada can support paragraph 4 as drafted, although our preference would be to phrase the opening sentence with "should refrain" instead of the current "shall". Mr. Chair, we recognize that the concept of context-appropriate human judgment and control has largely been discussed and that there is a range of views in the room with regards to this concept. My delegation agrees that while context appropriate human judgment and control is not an IHL obligation we do recognize the necessity for context appropriate human involvement which may include elements of control as a measure to ensure compliance with IHL. Having reflected on earlier comments we believe that box 3 may not be the most appropriate location for this concept since box 3 outlines proposed obligations. While we're still reflecting On a possible solution, one proposal could be to introduce the concept alongside other non-obligatory measures rather than an obligation. Finally, Mr. Chair, Canada wishes to emphasize the importance of consistency with IHL language and strongly encourages the use of well-understood terminology to describe IHL obligations. In paragraph 7A, the text addresses the concept of ethics and ethics considerations. Canada supports the notions that weapon systems must remain consistent with the principles of IHL, including the principles of humanity. Therefore, in para 7A, we see value in replacing ethical to reflect that functional effects should have due regard for principles of humanity. Similarly, in paragraph 7, which refers to the limitations of the operations of laws, as other delegates have said, we would suggest replacing the reference to military object with military objectives to be consistent with IHL treaty language. Thank you.

I thank the distinguished representative of Canada for sharing the views from his delegation and commenting on the text and making proposals. Next on my list is Israel, to be followed by Norway. Let me first pass the floor to the distinguished representative of Israel. You have the floor.

Thank you very much, Mr. Chair, and good afternoon to all of you, colleagues. I will share with you the comments on box 3. We have prepared them as detailed as possible, so it's a little bit long, so please bear with me. It's in the interest of being as clear as possible. Firstly, with regards to the division between the boxes, it is our view that box 3 should focus on the fundamental legal prohibitions and restrictions in IHL applicable to laws, while box 4 should focus on measures to implement those prohibitions and restrictions. Regarding paragraph 2, in our view, it is essential that the paragraph accurately reflect the obligation in Article 51 of the First Additional Protocol to the Geneva Conventions. Therefore, we suggest deleting the word "anticipated." The anticipation of the effects of an attack is not a primary obligation under IHL, but rather a second order requirement intended to fulfill certain primary obligations. And box three focuses on primary IHL obligations. We believe that the current text, as it stands, introduces a new legal standard, which is problematic. IHL requires to make a reasonable effort to try to assess collateral damage based on information reasonably available prior to the attack. as opposed to anticipating it with certainty. Unpredictable collateral damage does not render an attack as illegal. The current wording does not reflect those important nuances and may be understood as prohibiting many existing weapons or uses of weapons. The core point of anticipation of effects does not have a possible legal implication on the legality of weapons. is sufficiently covered by the reference to the possibility of limiting the effects of the use of a weapon. We are open to referring to anticipation of effects as a practical measure in the implementation of IHL obligations in a different part of the document, but not as it currently appears, namely as an independent legal requirement that by itself can disqualify weapons. This has no basis in IHL. Regarding paragraph 3, we recommend aligning this paragraph with the original language of article 51 of the Additional Protocol that this paragraph draws from, in order to avoid discussions on the implications of the deviations between this paragraph and article 51's language, and to ensure clarity regarding the legal obligation. Therefore, the paragraph will read: "The civilian population as such, as well as individual civilians, shall not be object of an attack by LOS." Regarding paragraph 4, similar to the request of other delegations, we asked to delete this paragraph, as in our view, the text should maintain its focus on the use of LOS and not other actions regarding it, for example, manufacturing or stockpiling. as they stray from the CCW's objective and expertise as an IHL-focused treaty. Regarding paragraph 6, the term "content appropriate human judgment and control" remains a serious concern to us. And we share the view expressed by the United States and the Republic of Korea on this issue. We believe it does not assist states in implementing the law. Furthermore, the context appropriate element can be easily understood as referring to an obligation of a constant real-time control, which is a controversial concept. For these reasons, we will request the deletion of this paragraph. However, if it is retained, for the purpose of getting closer to consensus, we are willing to positively consider the formulation suggested by the US delegation yesterday, meaning measures including during development, relating to control and human judgment are needed to ensure the use of force is in compliance with the principles and requirements of distinction, proportionality and precaution in attack and other rules of international law applicable in armed conflict. Regardless of its final exact wording, we will not be in a position to agree to the retention of this paragraph in box 3. In our view, a paragraph of this nature should not be appearing in box 3 since its focus is on primary obligations of IHL. Human-machine interaction is not a primary role of IHL, but rather a tool to implement the IHL. Therefore, we do not consider this box the appropriate location for this paragraph. Regarding paragraph 7. As we mentioned in our view, box 3 is not the appropriate location for this paragraph, since human-machine interactions is not a primary rule of IHL, but rather a tool to implement IHL. This paragraph lists a variety of practical measures, some of which may be needed to implement IHL in certain circumstances, and some that reflect policy measures that exceed legal obligations. Considering the practical nature of this paragraph, the appropriate place for it is box four, which outlines the measures that support compliance with IHL. However, should the paragraph remain in its current location, we believe several structural adjustments are necessary. First of all, it is important to clarify that this list is not a list of cumulative requirements, but rather of practical measures. that may be used in specific circumstances to assist in facilitating compliance with IHL. The practical or legal need for them will be context-dependent. From a legal perspective, the measures here are not legal obligations in and of themselves. And even in a situation where only some or even none of the measures are applied, it may still be possible to use laws in a manner that complies with IHL, depending on the circumstances. Furthermore, practically, we note that there are many -- sorry, that there are many -- there may be many -- I'm sorry. Practically, we note that there may be other measures that can be applied to enhance compliance with IHL, and that this is merely a list of potential examples. Secondly, we asked to delete the phrase "context-appropriate human judgment and control" for the same reasons we laid out previously. Therefore, instead of the current language of the chapeau, we will suggest to read: "Such measures can include, where appropriate, the following". Let me touch, please, Chair, about the specific sub-paragraphs of paragraph 7. Regarding paragraph 7, we find the coupling of assessment of legal and ethical considerations with the concept of responsible chain of command and control to be confusing. It mixes together too many concepts and terms. For this reason, we think that in this subparagraph, only the first sentence should remain. Regarding paragraph 7, in a similar vein to our comments to box 3, paragraph 2, We think it is important that this paragraph as well accurately reflect the obligation in Article 51 of the first additional protocol of the Geneva Convention. Therefore, we suggest replacing "control" with "limited". Furthermore, if this paragraph is moved to box 4, which does not focus on primary illegal IHS obligations, we think we can retain the word "anticipated" as it includes the qualifier adequately. Additionally, we believe that the words "for example" should be added, or alternatively, items from the list should be deleted or qualified significantly. as the examples given are only practical tools that may be used in specific circumstances to ensure the effects of loss are adequately anticipated and limited, rather than a mandatory requirement in every situation. Furthermore, there may be other measures that can be applied, and this is merely a list of potential examples. Therefore, the text of the paragraph will read, "Ensure the effects of laws can be adequately anticipated and limited by those responsible for their use, including, for example, by limiting or otherwise restricting the scale of the operation, which includes the types of targets, duration, geographical scope, and the number of engagements that LOS can undertake autonomously. Regarding paragraph 7C, we suggest deleting the language "without human intervention, including through the real-time machine learning" and replacing it with "in a way that will violate IHL." Conditioning compliance of IHL on the absence of human intervention, including through real-time machine learning, is problematic, as lawful systems updates may occur within predefined legal and technical constraints without requiring real-time human involvement during the operation. For instance, the system can be designed to operate within a predefined mission parameters and constraints, including limitations of the modification of target categories. Thus, we believe that the current wording misses the ultimate goal here, which is the prevention of violations of IHL. We suggest changing the text to clarify this. And to finalize, regarding paragraph 7, sorry, no, then I have paragraph 8, but this is the last two. Regarding paragraph 7, firstly, the requirement to limit the operation of laws to a defined perimeter is redundant, as geographical limitations are already addressed under section 7. Secondly, the reference to restricting the use to military objects seems to be misquoting or confusing IHL terms and rules. IHL does not require to limit attacks to military objects, but rather to military objectives. This is a mandatory obligation rather than a measure to ensure compliance, and presenting it in this way undermines the well-established prohibitions under IHL. In addition, this paragraph seems to conflate between distinction and precaution under IHL, as it does not offer added value in relation to paragraph 7 and entails other problems. We suggest deleting paragraph 7. Finally, regarding paragraph 8, We believe this paragraph should be moved to box 4, as it presents a practical measure, the consideration of IHL rules when developing a weapon, in order to maximize the chances that when the weapon is used, it will comply with IHL. Thank you.

I thank the Dutch delegation. representative of Israel for sharing the views of his delegations and commenting and making proposals. Next on my list is Norway, to be followed by Utrecht University and the ICRC. But let me first pass the floor to the distinguished representative of Norway. You have the floor.

Thank you, Chair. Box 3 operationalizes the two-tiered approach with elements of prohibitions and regulations, and this is a very important part of our text. Regarding paragraph 4, the first element is a commitment to refrain from manufacturing and stockpiling laws that it would be unlawful to use. There are some more elements, but those are the ones that we believe are the most important. In our opinion, that is an element that would add value to a future instrument. When large stockpiles of prohibited weapons exist, there is always a risk that they may be used, despite legal restrictions. Limiting the mass production and stockpiling of those helps to reduce that risk. We recognise that the classical notions of production and stockpiling are rooted in the experience of traditional, physically discrete weapon systems and may not translate very well to many types of laws. By contrast, many types of laws are likely to be distributed in nature, with components that are not physically co-located. and with development processes that are not linear. So, for instance, hardware and software and backend computational functions may be designed and updated or replaced at different times by different actors. And if you have software updates closer to deployment or changes to a remotely located AI component, for instance, that could fundamentally alter how the system functions. So in those cases, the classic model of regulating production and stockpiling might not be fully effective. because the core risk may not lie in the hardware itself. However, for some conceivable laws, we believe that that logic remains relevant. More broadly, we think that establishing a norm against mass-producing or stockpiling systems that one has deemed to be unlawful to use, for instance, pursuant to article 36 reviews, would send an important signal. We would like to retain the first part of paragraph 4. But if it could help to build consensus, we believe the first sentence could also be restricted to mass production and stockpiling. However, like others, we propose also to delete the second part of the paragraph. We recognize that it may be considered necessary in certain cases to construct a prototype or something in order to understand the system and explore possible countermeasures. Yet, as we see, there is no need for an explicit exception for this purpose. The operative focus of the first sentence is on preventing large-scale production of systems that a state has determined to be unlawful, and that does not bar the creation of some prototypes for defensive research and training. We think that, by including a dedicated exception, we risk normalising the existence of such systems, which is problematic. On paragraph 6, For us, this is a crucial element of the text for reasons that we have set out many times before. In short, we agree fully that context-appropriate human judgment and control is essential to ensuring that the use of laws complies with international humanitarian law. IHL requires a number of context-dependent judgments, and if those cannot be made or if their outcomes cannot be meaningfully implemented through appropriate levers of control over the system, then compliance with the law cannot be guaranteed. Of course, the reason we're talking about human judgment and control relates to the special concerns that laws raise, which classic weapon systems do not, regarding the potential that come with autonomy to outsource elements of human decision-making over life and death to machines. And just for the record, for us, the context that we're talking about here encompasses that range of relevant factors that are recognized in Guiding Principle C, including operational context, characteristics, and capabilities of the system, and so forth. and otherwise we think that the format you propose here, Chair, with the concept itself in paragraph 6 and then fleshed out with various possible measures in paragraph 7 is useful. Regarding subparagraph a, we're generally okay with it. For us, it's important to keep the sort of ethics hook somewhere in the text. In order to avoid the concerns around the words "due regard", we join others in proposing to change the last part of the paragraph to read "as well as assessment of relevant ethical considerations". Subsection contains many important measures. We support the contents, but, like others, we do not think that types of targets are part of the scale of the operation. We also support deletion of the words "which includes" in the middle. Of course, the modality that we choose or how we frame these measures depends a bit on how the chapeau ultimately reads, if it has the verb "should" or the verb "must" in it. But, the way the chapeau reads now, we believe that the current formulation in 7B is good. On C, We appreciate this subparagraph, which is an important measure to ensure that the weapon system keeps operating within the boundaries that have been set by the command. In our opinion, we should nevertheless change "human intervention" to "human authorisation", as has been mentioned by a few others as well, because a human intervention could be very trivial. The point here is that the system should not be allowed to make such significant changes without the express approval of a responsible human, and in our opinion, authorisation conveys this in a clearer way. If you authorise something, you assume the responsibility for it and can be held accountable. We also agree with Singapore and others that we should flesh out what these mission parameters are to improve on clarity. Of course, this is just elements for a prospective instrument that we may flesh out and negotiate in the future, and not something that we are intending to operationalize directly. But the more clarity we can add now, the better. Subparagraph e, as we understand it, this is an elaboration of some of the limits that are mentioned in subparagraph b, geographical scope and types of targets. And we believe this is a good risk-reducing measure to recommend that states undertake Now, some have questioned the term military objects, since that is not a word pair that we find in current sources of IHL, as opposed to civilian objects, for instance. If we want to stick to familiar IHL terms, at least for that part of the problem, we propose to draw some inspiration from the CCW protocols, for instance, number 3, article 1, and instead write objects that are military objectives. Finally, regarding the concepts of reliability, predictability, explainability and traceability, we regret that they have disappeared from the text, although some notion of predictability and reliability are captured by paragraph 7 on the ability to anticipate the effects of the weapon. As we have said before, we accept that some of these terms are more relevant to context-appropriate human judgment and control. and hence to box 3 than others, namely predictability and reliability. Traceability and explainability, on the other hand, are more relevant, possibly to the issues of accountability and responsibility. So I think Ireland elaborated this point very well. So like them, we believe that the first two should be reintroduced in box 3, whereas traceability and explainability should be considered for inclusion in boxes 4 and 5. Thank you.

I thank the distinguished representative of Norway for stating positions of his delegation and making comments and suggestions. Next on my list is Utrecht University, to be followed by the ICRC, Uruguay, and Pakistan. But let me first pass the floor to the distinguished representative of Utrecht University. You have the floor.

Thank you, Chair. I'm again intervening on behalf of the Asia Pacific Institute for Law and Security delegation and my own Utrecht University. We welcome the change in paragraph two. The requirement for the effects of the attack to be anticipated and limited is more in line with existing IHL, especially Article 51 of Additional Protocol I. In paragraph three, we note the deletion of the phrase "in all circumstances". Admittedly, the deleted language follows existing treaty law, such as Article 2 of the CCW Protocol III, which provides that it is prohibited in all circumstances to make the civilian population as such, individual civilians, or civilian objects the object of attack by incendiary weapons. While we do not wish to undermine the protective intent of this type of language, we recall that under IHL, civilians lose protection from attack for such time as they directly participate in hostilities. The inclusion of the phrase "in all circumstances" can therefore be read as departing from the established IHL framework. Therefore, we support the deletion of this phrase because this allows the customary international law rule regarding direct participation in hostilities to operate undisturbed. We do, however, suggest that the wording be altered slightly to reflect existing language in Article 51 of AP I, The civilian population as such, as well as individual civilians or civilian objects, shall not be the object of attack by laws. Chair, paragraph four is a further elaboration of the language that was previously included in box four nine. We agree that box three is a more appropriate location for this paragraph. But to avoid interpretive challenges, we suggest aligning the language more closely with the wording used in recent arm controls instruments, especially the anti-personnel mine ban convention the Convention on Cluster Munitions, as other delegations have mentioned. Thus, the first sentence of paragraph 4 should also cover deployment and production and both direct and indirect transfer. Noting that the use of laws is already addressed elsewhere, the sentence could read as follows: "States undertake not to develop, produce, otherwise acquire, stockpile, retain, or transfer to anyone, directly or indirectly, laws that cannot be used in compliance with IHL." As other delegations have mentioned, we are not entirely convinced of the need for the exception in the second sentence. We would be interested in hearing the views of delegations as to whether laws that cannot be used in compliance with IHL require such unique countermeasures that is necessary to possess unlawful laws just to develop and test these countermeasures. If this is indeed the case, coherence in language with other arms control treaties is desirable. As we highlighted in our box 2 intervention yesterday, for internal coherence of the text in paragraph 6, we recommend the following addition: "Context-appropriate human judgment and control is needed across the life cycle of laws." This would bring the text in line with existing language in box 4, paragraph 3, and ensure consistency throughout the text. Chair, with respect to paragraph seven, we are concerned that removing the clause in order to uphold compliance with IHL, which was previously the beginning of the chapeau, loosens the connection between the measures listed in IHL compliance. We recommend retaining this reference to IHL in the chapeau. We appreciate the reorganization, consolidation and streamlining that has occurred in the subparagraphs. We note that a number of delegations have expressed this concern about the omission of concepts of predictability, reliability, traceability and explainability. As we have previously pointed out, the substance of many of these concepts has been incorporated into various parts of the rolling text. So therefore, we think notions such as predictability and reliability may deserve explicit mention here, following on from the Norwegian delegation just now. whereas traceability and explainability are encapsulated in the current wording of box 5. In subparagraph, we welcome the explanation of the meaning of the term "scale of operation" through a number of examples. We agree with delegations that have supported the retention of this list of examples. However, "limiting or otherwise restricting" seems to be a linguistic redundancy. We struggle to think of a restriction that would not also be a limitation. With regards to subparagraph C, we believe more specificity is required regarding the parameters mentioned. Additionally, the word significantly may be overly ambiguous, and the idea of a system modifying itself tends to anthropomorphize technology. We suggest a rewording of the subparagraph that reflects cohesion with the rest of the text, including with reference to legal review requirements as outlined in box four, and suggest the following. ensuring that mission parameters do not change without context-appropriate human judgment and control, and any further development or modifications to laws that have an impact on their functions or anticipated effects must be reviewed for compliance with IHL. We are uncertain about sub-paragraph, which seeks to restrict the use of laws only against military objects. In the previous text, subject paragraph D4 sought to limit the use of laws to attacks against objects that are military objectives by nature. The new language in paragraph E is therefore more restrictive and has the effect of prohibiting the use of anti-personnel AWS. We are not sure if this was the objective of the drafters. If it was, we believe the ICRC's proposal to place such a prohibition into a dedicated paragraph would add clarity. Finally, in paragraph eight, foregoing might be a better word than formulated when referring to the requirements and measures contained in paragraphs one through seven. And we suggest the wording, the foregoing prohibitions, requirements, and measures must be taken into account across the entire life cycle of laws for consistency with the rest of the text. I thank you, Chair.

I thank the distinguished representative. of Utrecht University, and let me now pass the floor to the ICSC to be followed by Eurogamer, but let me first pass the floor to the distinguished representative of the ICSC.

Thank you, Chair. Echoing the comments made by the distinguished representatives of Spain, Switzerland, New Zealand, Mexico, Austria, Sweden, Panama, and Ireland, among others, we welcome the reference to ensuring context-appropriate human judgment and control and the use of laws, as we see in the chapeau of paragraph 7. It is important to emphasize that the current language already reflects significant compromise that was done in the spirit of consensus building. We recall that this discussion began with references to meaningful human control, which later evolved into considerations regarding nominal human input and subsequently into the formulation of context appropriate human judgment and control. We also recall that the presence of this concept initially in box two was originally not accepted by some delegations as it was seen as too ambiguous. In our view, the inclusion of this requirement in paragraph seven and tying it to concrete restrictions and measures illustrate precisely what is meant by this concept. Rather than prolonging debate over terminology, this approach anchors the concept as a guiding principle and then sets out restrictions in subparagraphs A to E to illustrate what this entails in practice. Now, with regard to subparagraph 7, we recommend deleting the first clause of the paragraph and beginning the sentence directly with "limit or otherwise restrict." This would more accurately reflect the purpose of the second tier of the two-tier approach, which is to articulate concrete restrictions on how autonomous weapon systems are used. In addition, several of the restrictions listed serve purposes that go beyond ensuring the effects of autonomous weapon systems are anticipated and limited. For example, restrictions relating to types of targets directly operationalize the principle of distinction, which is concerned not only with effects, but with the prior determination of lawful targets. As such, framing paragraph 7 in this way would better reflect the breadth of relevant IHL considerations and align more closely with established conduct of hostilities principles. Also, echoing the comments of the distinguished representative of the US, we also recommend deleting the term "which includes". For as they eloquently put it, the scale of an operation does not necessarily encompass all elements listed after, in particular, the types of targets. As such, in our view, removing these words would avoid conflating distinct form of restrictions and improve the clarity of the paragraph. Drawing on language from paragraph 60 of the May version, this paragraph could thus be amended to read as limit or otherwise restrict the scale of operation, types of targets, duration, geographical scope, and the number of engagements that laws can undertake autonomously. On 7C, echoing comments made by the distinguished representative of Singapore, we find mission parameters to be ambiguous as well. Now, drawing on the wording of paragraph 7C of the May rolling text, we recommend adding the phrase, "With regard to the types of targets, duration, geographical scope, and scale of the operation, which cannot be modified." following the reference to mission parameters. This way, it could read as, "Ensure that mission parameters with regard to the types of targets, duration, geographical scope, and scale of operation of laws," and it goes on. This would better link paragraph 7 to the previous paragraph, clarify the meaning of mission parameters, and avoid the use of different terms to describe the same concepts, which would enhance legal clarity and coherence. In addition, echoing the remarks made by the distinguished representatives of Panama, Colombia, New Zealand, Mexico, Austria, and Sri Lanka, among others, we recommend deleting the term significantly as it may unduly narrow the scope of the paragraph and lacks clarity as to the threshold of modification that would trigger its application. Moreover, as supported by many delegations in previous sessions, we recommend replacing the phrase "without human intervention" with, without context appropriate human judgment and control. In our view, without human intervention risks allowing nominal or formalistic human involvement to be considered sufficient, whereas the latter formulation would ensure that any modification of mission parameters, system functions, or effects remain subject to contextual human judgment and control. It also ensures consistency throughout the text. We further suggest that paragraph 7C explicitly refer to the legal review obligation mentioned in box four of paragraph two to clarify that modifications to functions or effects of autonomous weapon systems would also trigger the need for a legal review to ensure compliance with IHL. We also recommend deleting the phrase, "including through real-time machine learning." While we appreciate the intention to ensure that autonomous weapon systems with machine learning capabilities are addressed, We recommend using a technology neutral formulation so that the rolling text and the broader work of the GGE remain relevant in light of future technological developments. Moving on and in line with our comments earlier, we propose adding a reference to predictability, reliability, traceability and explainability as a separate sub paragraph. As a point of brief clarification, we refer to paragraph 21 of the 2023 GGE report. To be precise, that formulation appeared in earlier draft iterations of the report and reflected elements that received broad support during discussions, but it was not retained in the final adopted text. That said, our point remains that these elements were extensively discussed and supported by a number of delegations in the GGE process, including in today's session. We therefore continue to consider their inclusion in the present text to be both relevant and quite necessary to ensure compliance with IHL. To briefly elaborate, autonomous weapon systems which function in opaque ways, such as those relying, for example, on non-deterministic AI systems, effectively prevent the human user at the time of the weapon's legal review, at the time of its use, or both, from being able to understand and predict the system's effects. This impossibility prevents the user from being able to direct it against a specific military objective and effectively results in a lack of control over the weapon's effects, making it indiscriminate by nature. More generally, even where an autonomous weapon system is not considered indiscriminate by nature, users who employ an autonomous weapon systems whose effects they cannot sufficiently understand and predict in the specific circumstances of use would violate the prohibition of indiscriminate attacks. This is because in those circumstances of use, employing the system would be akin to blind firing. Alternatively, the inclusion of those terms could be done in line with suggestions made earlier by the distinguished representatives of Ireland and just now by Norway and Utrecht University. Finally, with regards to paragraph 7e, we understood the reference to military objects as referring to objects that are military objectives by nature, a concept that we and other delegations have supported as an important safeguard for minimizing the risk of harm to civilians and civilian objects. Should this be changed to military objectives, as has been suggested by a few delegations, then we would recommend reverting to the previous version of the rolling text, which distinctly outlined two restrictions in previous paragraph 6D, 3 and 4, namely, limit the operation of laws to a defined parameter, in particular to reduce the likelihood of harm to civilians or damage to civilian objects. And the second one would be restrict the use of laws to objects that are military objectives by nature. I thank you, Chair.

I would like to thank the distinguished representative of the ICSC for commenting on the text and making proposals. Next on my list is the distinguished representative of Uruguay. You have the floor, Madam.

Thank you, Chair. First, I want to thank for the wording of paragraph 3, which we think offers a decent basis to move towards more defined normative elements. We support the text presented for paras 1 to 6. As with previous speakers, we think that it is fundamental to ensure context-appropriate human judgment and control in the development of such systems, and we support moving it from box 2 to box 3 and linking it to IHL. Also, we welcome the compromise solution, taking into account the various positions expressed so far. In 7C, we share the concerns expressed by other delegations for using the word "significantly", and we would be in favour of deleting "significantly". On paragraph 8, we believe that the whole life cycle is taken into account and that it should take out "requirements and measures". We would delete "requirements and measures".

I thank the distinguished representative of Uruguay for commenting on the text and making proposals. Next on my list is Pakistan, to be followed by Belgium, Finland and the University of Southern Denmark. Sorry. But let me first pass the floor to the distinguished representative of Pakistan. You have the floor, sir.

Thank you, Chair. Chair, my delegation now turns to the remaining paragraphs of box 2. We had already addressed up to the paragraph -- from paragraph 1 to paragraph 5. So from paragraph six onwards, I'd like to share our position. And in doing so, I would also like to comment on some of the proposals that we have heard from the floor this morning and this afternoon. So I would like to request you and seek your indulgence because of the length of this intervention. Chair, Something that we have heard consistently through this session and previously as well, and that really led to highlighting of our position also, that what we are doing here is not just restatement of IHL. And if our task was just restatement of IHL, we believe we would not have needed the three-year mandate of the GGE. We would have been able to, I think, efficiently do it and finish and conclude our work in one year. So, Chair, we continue to approach the work with that conceptual position. On paragraph 6, we think that this is the crux of our entire text, the issue of human control. It would need to be resolved one way or the other. We have heard this argument that this is not an established concept in IHL treaties and it should not be introduced here. We have noted previously also that there is a reason why we do not find this term. And when we listen to such arguments, I think they confuse terminology with the substance. Until the emergence of lethal autonomous weapon systems, the human control over the critical functions of targeting, identifying, selecting, and engaging was always inherently assumed. There was no need to spell it out. It was always inherent. I think laws, in our view, are the first category of weapon systems where this fundamental assumption breaks down. And that is precisely why we now face the need of articulating it, which was always presumed earlier. Chair, we have also heard the argument, including today, that human judgment and control should be treated as means rather than ends. And also, we heard that automation and artificial intelligence are legitimate means and methods of warfare that can make the use of force more precise, and that a human control standard risks stigmatizing innovation and privileging human involvement over civilian protection. We wish to address this argument directly because we think it reflects a false choice. Chair, no delegation in our view, and certainly not ours, is arguing that human control is an end in itself, divorced from civilian population, quite the opposite. We are arguing that human control is the mechanism through which civilian protection is achieved under IHL. The obligation to distinguish, the obligation to assess proportionality, the obligation to take precautions These and other obligations are all that fall on humans. They require human judgment, precisely because they involve the kind of contextual, qualitative reasoning that determines whether civilians live or die. Chair, we fully accept that technology can enhance precision. We fully accept and see the logic that automation can, in certain circumstances, reduce risks to civilians. But that is an argument for well-designed systems functioning with safeguards that we are trying to establish here, operating under human judgment and human control. Chair, we also wish to make further observation that we note with great interest that some delegations who resist the general formulation of context-appropriate human control and judgment in the chapeau, for instance, nonetheless are able to support many of the specific measures in the sub paragraphs that follow. For instance, a responsible chain of command, the ability to anticipate control and effects. Some have even proposed reframing these measures as feasible precautions in attack. But what are these measures, whether we call them regulations or precautions, if not concrete expressions of human control? Each one requires a human to set parameters, anticipate effects, retain the ability to intervene, and limit autonomous operation. In our view, if we can agree on specific measures, we think we should be able to agree on the general principle that is able to capture these individual elements. Finally, on this issue, some words on the qualified context appropriate, since we have also heard the concern that it is too vague. a policy standard rather than a legal one, and it reduces the range of relevant factors to context alone, ignoring system characteristics. Thank you. So, I'd continue. We have heard this concern that context is too narrow and it only captures certain notions. We do not see, we do not read the qualifier so narrowly. Context in operational usage in our view, encompasses exactly those factors which we have heard in the room, the nature of the system, the environment, the capabilities of the operator, the characteristics of the target set. And when we look at the guiding principles, see that we adopted in 2019, from which this language in our view is drawing upon. We had spoken of, and I quote, "operational context and system characteristics," unquote, together. In our view, the word context is capacious enough to encompass all these circumstances. And we recall that all of IHL's core concepts, like military necessity, proportionality, involve context-dependent assessment. We are open to the idea of further spelling it out if it helps in building convergences on this. But as of now, we certainly feel that we would like to retain this concept as it appears right now. Chair, now let me move to the issue of predictability, reliability, traceability, and explainability. We note with concern that this language no longer appears in the text as it used to in previous iterations. And we have heard today also, we've heard in the past that these are not part of existing IHL. text, and we disagree with that. We think that when we look at the obligations under IHL, distinction, proportionality, and precautions, these are legal duties. But we fail to see how these duties can be discharged by an operator or a commander if a system behaves unpredictably, unreliably, or in a way that cannot be explained. If a commander cannot foresee how a system will act cannot rely on it to act consistently, then we fail to see that how they can understand why it made a particular targeting decision during operations. Chair, we do not use these metrics, we do not use the metric of whether these words appear in IHL before or not. We think these are 21st century safeguards for 21st century challenges. And the qualifier that you had in the previous text adequately. We think it provides necessary context dependence that we are not seeking absolute predictability, but a level sufficient for IHL compliance. Chair, on sub-paragraphs, on paragraph 7, on the chain of command reference, we have heard the suggestion that perhaps it belongs in some other boxes on accountability grounds. We are open to exploring that, provided the concept is adequately captured anywhere in the text. However, we would like to note that the chain of command is not only linked with accountability mechanism. It is the operational structure in our view. It is the operational structure through which, for instance, precautionary obligations that are set in IHL are discharged. Chair, in paragraph 7B, The requirement that effects can be adequately, should be adequately anticipated and controlled has our strong support. We urge for its retention. We think that the standard of anticipating effects is already embedded in Article 57 of Additional Protocol I, and this sub-paragraph merely applies the existing standard to lethal autonomous weapons systems. It's not a new standard. On the number of engagements, limiting autonomous engagements before requiring renewed human input, we think it's a necessary precautionary measure. It constrains the consequences of errors and ensures periodic reassertion of human judgment. This is also consistent with the Article 57 obligation to cancel or suspend an attack, for instance, when it is becoming disproportionate. We note with interest the proposal, for instance, to reframe these measures as feasible precautions in attack, and we can engage constructively on it. But many of these measures are indeed precautionary in nature. But the framing that we choose, it should not become a vehicle for weakening the measures themselves. So we will be engaging on the language in that spirit. Chair, on 7C, We support the principle that mission parameters cannot be significantly modified by the system without human intervention. We acknowledge that not all modifications are operationally significant. Technical adjustments to flight paths, for instance, do not affect targeting, and we are open to language that focuses the restriction on modification with IHL implications. On real-time machine learning, our concern is not with the systems that refine their ability to recognize predefined target categories to improve safety, for instance. We heard today an argument that it could be useful for enhancing precision, the real-time machine learning. Our concern is that in case of systems that autonomously redefine what they are authorized to engage, altering entire target categories or engagement criteria during a mission without any human intervention. And there is a meaningful distinction in our view between optimizing within defined parameters and overriding those target parameters altogether. And we think that that can be fixed. We have heard proposals, for instance, from the delegation of Brazil today as well. We are open to working on the language. We can add specific focus on target selection and engagement function in this paragraph to make it clear what we want to aim at. Lastly, Chair, on paragraph 8, we have heard two concerns. One, that the phrase "must be taken account" is too vague, and second, that IHL does not regulate development. On the vagueness argument, Chair, we are open to strengthening the language. For instance, we think that it could be changed to "must be respected" in the use of laws or "must inform their design and development". In our view, the obligations applicable to use must be anticipated at the design stage. And we have made this argument at length in our previous meetings. We think that Article 36 of AP1 and our previous agreements in guiding principles that IHL applies during the development and use of laws, it necessitates the inclusion of the development phase of these measures. We are sympathetic to the suggestions that the paragraph can be specified, which provisions are we referring to, and we are also open to including references to training and legal reviews along with design, development, and use. I thank you.

Let me thank the distinguished representative of Pakistan for stating the positions of his delegation, reacting to previous interventions and making proposals. Next on my list is the distinguished representative of Belgium, to be followed by Finland. Let me first pass the floor to the distinguished representative of Belgium.

Thank you, Mr. Chairperson. As this is the first time that my delegation takes the floor, 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. Mr. Chairperson, 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 accumulation 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 maintained. In this regard, I would like to echo the remarks made by Switzerland that those delegations, including our own, that strive for a wide scope and a strong framework have already made substantial concessions. It is such flexibility that allows our deliberations to move forward, and we call on all delegations to approach this GGE in such a manner. Mr. Chairperson, with your permission, allow me to briefly address the first two boxes. Belgium wants to avoid unnecessarily restricting the scope of the characterization. We favour a wider scope, as the two-tier approach later on, in any case, distinguishes between those laws that are to be permitted and regulated and those that are to be prohibited. We support the text as proposed in box 2 and would like to see it maintained. Mr Chairperson, this brings me to box 3. In Belgium's view, overall, the text in this box is in good shape. However, we have a couple of remarks. Let me start by echoing the comments made by Ireland earlier on the removal of the reference to predictable, reliable, traceable and explainable. We consider these concepts valuable in the light of our work, and we would like to see them reintroduced in the text, either in their original form, but we are also open to considering the alternatives referred to by Ireland. On paragraph 4, we see risk in the language proposed regarding the retention of laws. Unlike anti-personnel mines or other tangible weapons, laws are far more scalable due to their composite nature and encompassing software. Therefore, my delegation would like to see the second sentence of paragraph 4 removed. happy to see a provision on real-time machine learning reflected in the text under paragraph 7. That was originally one of our suggestions. As we see it, the aim is not to prevent machine learning. It makes sense that the laws learn from operations to improve their functioning. However, we want to avoid that the laws reprograms the mission objectives or the main modalities of its critical functions during the mission itself. This would lead to an unacceptable level of unpredictability. Such a scenario could feature friendly fire or harm to civilians, just to give two examples. Having said that, to better clarify this notion, we would suggest adding the wording, "while active in an operational context," before comma including, if considered useful. We are also open to consider the Austrian proposal in this regard. Furthermore, we consider the term significantly unduly restrictive in this context and would favor its removal. We would be open to substituting the term "human intervention" in this sentence with "context-appropriate human judgment and control" to better capture the role of the human element, as suggested earlier by the ICRC. Mr. Chairperson, let me conclude by stating that we value the reference to the design and development phases under paragraph 8 and would like to see these retained. I thank you, Mr. Chairperson.

Let me thank the distinguished representative of Belgium for commenting on the text and making proposals. Next on my list is Finland, to be followed by the University of Southern Denmark, then the University of Nottingham, Bulgaria and India. But let me first pass the floor to the distinguished representative of Finland. You have the floor.

Thank you, Chair. Box 3 has developed in many respects in a positive direction. The consolidation of elements related to context-appropriate human judgment and control into the same section is a welcome improvement. There is still some room to refine the structure of the section. For instance, paragraph 5 could be placed at the very beginning of the section for improved coherence. Regarding paragraph 4, some clarification is needed to ensure that its formulation does not contradict IHL. In particular, we find the reference to acquisition problematic, especially with regard to Article 36, and recommend it to be deleted. However, we can, to some extent, understand the rationale behind the addition of the last sentence of this paragraph. namely that the future instrument should not hinder the development of countermeasures against loss. For example, in a situation where a state would have to momentarily retain a prohibited type of loss as spoils of war, it should be able to do so for defensive research purposes. In this respect, we understand the comments made by the delegation of the Republic of Korea and could support the retention of the paragraph in some form. However, we have heard many delegations expressing their concerns on the last sentence of the paragraph and suggesting further conversation on this in the future negotiations. Therefore, we are also prepared to accept the deletion of the entire paragraph at this stage. With respect to paragraph 7A, it is regrettable that the formulation regarding ethical considerations has been weakened and we would have preferred the previous wording. Ethical considerations play an important role in helping ensure that all fundamental principles of IHL are complied with. Regarding paragraph 7c, we support the comments made by the delegation of Japan and would like to retain the term significantly. Concerning paragraph 7e, The prior reference to military objectives by nature has been replaced with a more general reference to restricting the use of laws to military objects. In line with the delegation of Singapore, Japan, and Spain, we suggest referring to military objectives as defined in IHL, including Article 52 of Additional Protocol I and relevant customary law. I thank you, Chair.

I thank the distinguished representative of Finland for commenting on the text and making proposals. Next on my list is the University of Southern Denmark, to be followed by University of Nottingham and Bulgaria. But first, the distinguished representative of the University of Southern Denmark, you have the floor.

Thank you, Chair, for giving me the floor. I'm speaking on behalf of AutoPractices, an international research project hosted by the Centre for War Studies at the University of Southern Denmark. We are grateful for the opportunity to observe the GGE's debates and share our expertise. Chair, distinguished delegates, please allow me to make some remarks regarding paragraphs 7 and 8 of box 3 of the rolling text, the paragraphs concerning measures to ensure context-appropriate human judgment and control. And these remarks are based on the findings of the recently concluded Auto Practices Project and its main outputs, published in collaboration with the Stockholm International Peace Research Institute. A toolkit of best practices and measures to sustain the exercise of human judgment and control across the life cycle of AI systems in the military domain. The best practices in this toolkit have been identified by 49 stakeholders who represent different disciplines and regions of the world. The first remark concerns paragraph 7 and the subparagraph A. Subparagraph A matches one of the main takeaways of the Auto Practices Toolkit, namely that the exercise of human judgment and control is strengthened by conducting continuous legal, ethical, and security assessments that are appropriate to the context of development and use. The Auto Practices Toolkit highlights that such assessments should be conducted throughout the life cycle of laws, and especially at the early stages, even prior to the development of weapon systems. The second remark relates to paragraphs B to E, which also match some of the conclusions reached by our project stakeholders. One takeaway is that the exercise of human judgment and control is strengthened by setting boundaries and limitations to the development and employment of laws. This is particularly important for the early stages of the life cycle, as whatever happens at those early stages affects the quality of human judgment and control later on, including at the use of force stage. Setting specific objectives and establishing boundaries for developing and using laws strengthen humans' understanding of how these systems work in particular contexts and humans' ability to reasonably foresee systems' impacts within those contexts. And boundaries and constraints can be spatial, temporal, as well as relating to the type of targets, the targeting parameters, or the operational environment, as suggested by subparagraph B. Auto practices stakeholders highlight that limitation and restrictions should reflect political, legal, ethical, and societal considerations surrounding the integration of AI systems into the military domain. These considerations should be informed by feedback loops and communication mechanisms from the end users of systems such as operators. And this takeaway on boundaries and limitations to laws also speaks to subparagraph C, D, and E of paragraph 7. Another of our project's findings is the importance of mechanisms ensuring human intervention, which matches subparagraph C on ensuring that parameters of laws are not significantly modified by the systems. As auto practices stakeholders highlight, intervention mechanisms that contribute to ensuring that laws are employed in ways that align with strategic, operational, legal, ethical, and other considerations applicable to the context, as well as the limitations set at the early stages of the life cycle, strengthen the exercise context appropriate human judgment and control. These mechanisms also contribute to establishing safeguards for laws to be employed in ways that do not supplant or replace human decision-making, which is particularly important for tracing accountability back to human actors along the life cycle. These mechanisms therefore reinforce subparagraphs a's point on responsible chain of command and control. Feedback loops and continuous communication between end users and developers of laws can be useful practices to assist in delimiting responsibilities and objectives of employing AI technologies in the relevant context. Risk assessment frameworks that are intended to provide benchmarks to ensure that laws are used for intended objectives are also a best practice highlighted by our project. Robust metrics for evaluating outputs and parameters of laws and their wide-ranging implications for operations can support relevant actors to engage in the intervention mechanisms mentioned previously. And risk assessment frameworks are closely connected to the sets of practices highlighted. For example, they may include intervention mechanisms. And they can also include post-use review mechanisms, highlighting the value of a life cycle perspective to retain context appropriate human judgment and control in the use of force. And finally, our project adopts a life cycle approach and includes sets of best practices at each stage of the life cycle of AI systems in the military domain. And this approach matches the message conveyed by paragraph eight in box three. the importance of applying measures and practices such as limitations, intervention mechanisms, feedback loops and risk assessment frameworks throughout the design, development and use of laws. At the same time, we support the representative of the University of Utrecht's previous suggestion to consider using the wording throughout the entire life cycle of laws, as in box four. as this would represent a more comprehensive perspective and take into account other stages than the design, development and use, such as the pre-development stage and post-use stage of the life cycle. Thank you, distinguished delegates, for your attention, and thank you, Chair.

Let me thank the distinguished representative of the University of Southern Denmark, and let me now pass the floor to the distinguished representative of the University of Nottingham, to be followed by Bulgaria, India, and Australia. I will start with the distinguished representative of the University of Nottingham.

My observations and recommendations are in relation to the legal interpretation and consequences of box 3 in its current format. The box is intended to be the main provision for regulating laws through prohibitions and restrictions. However, it confuses two distinct areas of law, namely weapons law and targeting law. Weapons law concerns the lawfulness of weapons their generic intended circumstances of use. Targeting law is on the lawfulness or otherwise of the decision to attack a target on a specified occasion using particular weapons in particular ways. By confusing those two areas, the current text makes it unclear what is prohibited and what is restricted. It only refers to use of laws which may be subject to prohibitions and restrictions and does not mention the nature of laws which may also be subject to prohibitions and restrictions under IHL. Under Article 35, paragraph 2 of Additional Protocol I to the Geneva Conventions and customary international law, IHL prohibits weapons which are by their nature causing or cause superfluous injury or unnecessary suffering. So there is no question of usage as the weapon is prohibited before it can be used. Therefore, laws which by their nature cause superfluous injury or unnecessary suffering are prohibited. Under Article 51, paragraph 4, subparagraph B and subparagraph C of Additional Protocol I to the Geneva Conventions and customary international law, IHL prohibits weapons which are by their nature indiscriminate because they cannot be directed at a specific military objective or because their effects cannot be limited as required by IHL. Therefore, laws which by their nature cannot be directed at a specific military objective or their effects cannot be anticipated and limited are prohibited. Taking these two legal points together, I would suggest the inclusion of the following text. I quote, "Laws which by their nature cause superfluous injury or unnecessary suffering, or which cannot be directed at a specific military objective, or their effects cannot be anticipated and limited, are prohibited. Coming to the use of laws, if the weapon is not prohibited, then its use must comply with the principles of distinction, proportionality and feasible precautions. I note there are concerns with the current formulation of paragraphs 5 and 6 and would suggest merging those two into an amended paragraph as follows: "If laws are not prohibited, their use must be in compliance with international humanitarian law. States shall ensure that human judgment and control is exercised in relation to principles of distinction, proportionality, feasible precautions open brackets, verifying targets, semicolon, choosing methods and means, semicolon, or assessing the likelihood of excessive loss, comma, injury, comma, or damage, close bracket, full stop. I have placed context appropriate in square brackets as its meaning is unclear at this stage, and it's yet to be agreed by state delegations. I think that the more relatable concepts of human judgment and control can be readily mapped into implementation of existing principles of distinction, proportionality and feasible precautions. I am happy to send that text to you to share.

Let me thank the distinguished representative of the University of Nottingham, and indeed, we would appreciate receiving your comments in writing. Next on my list is Bulgaria, to be followed by India, Australia, and the US. But let me first pass the floor to the distinguished representative of Bulgaria. You have the floor.

Thank you, Chair. We appreciate the streamlining of the formulations on paras 1, 2, and 3, as they constitute key IHO provisions. With regard to para 4, and in particular its second sentence containing the retention clause, we consider that this formulation is more closely associated with other disarmament instruments and categories of weapons and not suitable for the specific context of lethal autonomous weapon systems. A perspective shared and explained in detail by several delegations already. We welcome the placement of the new para 6 concerning context appropriate human judgment and control in box 3. We value this concept as an appropriate means of reflecting the broader notion of the human element. Implementing Context appropriate human judgment and control is key in ensuring compliance with international law and international humanitarian law during the development and use of laws. Accordingly, para 7 provides us with a list of practical measures. We have few concerns related to para 7, and more specifically to restricting the number of engagements in 7 and the formulation military objects and its inconsistency with IHL in 7. To our benefit, the delegations of Japan and Singapore made some suggestions to reformulate the language, thus making it more comprehensible and admissible.

Thank you, Chair.

I thank the distinguished representative of Bulgaria for stating the position of his delegation. Next on my list is the distinguished representative of India, to be followed by Australia, US, and China. But first, the distinguished representative of India, you have the floor.

Thank you, Mr. Chair, and good evening to all colleagues. First of all, we should thank you for your patient and very steady stewardship of our work. You certainly have an unenviable job given the diversity of perspectives that have emerged today in our discussions. We did not intervene earlier today as we were carefully listening to and reflecting upon the various views that have been expressed by delegations. Our view is that we should thrash these issues out in a systematic manner, perhaps by moving into an informal mode of conversation. Nevertheless, I would like to take the opportunity to formally place our views on Box 3. At a general level, my delegation approaches Box 3 in a constructive spirit. We support meaningful and progressive regulation anchored in IHL. We additionally recognize the value of Box 3 in clarifying how IHL applies to laws, including where appropriate, by elaborating concepts or measures that address particular concerns raised by these systems. At the same time, we believe that clarity and legal precision are essential to ensure coherence, avoid ambiguity, and maintain consistency with established legal principles. In this regard, we suggest careful attention to the words used in box 3. Where the text merely reiterates binding obligations already contained in IHL, the use of "must" would be appropriate. In other instances where the intention is to develop new measures, The choice of verb should be deliberate and reflect the intended legal effect. More specifically, we have the following views. We are open to accepting the prohibitions as formulated in paragraphs 1 through paragraph 3. We have heard suggestions to calibrate the language more closely to established IHL formulations for the sake of consistency. We support that suggestion and consider it both constructive and useful. On paragraph 4, we align with the view that IHL traditionally prohibits use rather than development as such. That said, we recognize the concern behind this provision and see value in ensuring that systems which are inherently incapable of lawful use do not circulate. However, we believe that some refinements are necessary on this paragraph. First, the reference to manufacturing may be redundant once stockpiling is addressed. Moreover, manufacturing could arguably affect production for legitimate research and development activities, including testing and evaluation. Second, and more importantly, the paragraph should clearly distinguish between, first, weapons that are inherently incapable of being used in compliance with IHL, and second, weapons that are lawful by nature but may be misused. Only the former category should fall within this prohibition. We therefore suggest clarifying that the restriction applies to systems that are by their nature incapable of being used in compliance with IHL. Given the concerns that we have heard, we are also willing to merge this with paragraph 8, which also has the same operative meaning, but may have a more palatable formulation. We are also willing to address the contents of para 4 and para 8 combined in box 4, if that is likely to gain consensus. My delegation would also like to propose an additional prohibition, if I may put it that way, which we previously shared during informal consultations. Allow me to briefly explain the rationale first. In our view, obligations under IHL bind states and not terrorists or non-state armed groups. Such actors, if they are to acquire such weapons, are highly likely to use these weapons recklessly, often with deliberate aim of carrying out indiscriminate attacks. They cannot be relied upon to comply with IHL, nor can they be effectively held accountable. Their use of these weapons would therefore raise serious humanitarian concerns. It is thus important to ensure that states do not transfer these weapons to terrorists or non-state armed groups. Accordingly, we propose the following language. Laws should not be transferred to terrorists or non-state armed groups. States must ensure proper tracing and record keeping of the weapon system throughout its life cycle. We are, of course, open to engaging further with other delegations on this proposal. We note that the proposal from our side has come late, but we are willing to engage further and also discuss a possible placement in this text. On para 7, we agree with some delegations of the need to streamline sub-paragraphs under para 7. The word "ensure" is repeated twice, once in the chapeau and the next time in each of these subparagraphs. We believe that this redundancy could be removed. More generally, the measures have not been presented in a uniform and consistent way. We request for edits in that regard. Regarding paragraph 7, we have taken note of the view that legal obligations and ethical considerations are distinct and, therefore, should not be conflated. That said, we support the inclusion of ethical considerations in the text. In our view, ethical considerations are important. We recognize that there is no universal benchmark for what constitute ethics in warfare, and that such considerations may potentially be subjective. However, even when they are subjective, the paragraph merely encourages states to show due regard to ethical considerations. This is both reasonable and appropriate in our view. At the same time, states retain full discretion to make engagement decisions as they deem appropriate after having given due regard to ethical considerations. In light of the concerns expressed, we would be also open to placing this paragraph in box 4, given its very nature, but our preference is to accommodate the reference to ethical considerations. In 7B, after the term "anticipated and controlled," we would like to insert "as required by IHL." This is how it has been mentioned earlier in paragraph 2 in this box. Also, we are comfortable with retaining limits on number of engagements in paragraph 7B, We note that IHL does not prescribe limits on number of engagements. States will, however, have the flexibility to define limits by themselves. Limiting the number of engagements can help reduce unintended engagements, for instance, in case an autonomous weapon system has malfunctioned, and it can help in mitigating humanitarian risks. If this notion is not acceptable in this box, we suggest preserving this notion in box 4. On 7C, my delegation has noted the changes you have made from the last draft. We believe that the previous formulation built around control and human judgment should have been used. This, in our view, would function also as a safety mechanism. In case a laws accidentally moves into a civilian concentrated area, the system should by itself be able to itself modify its parameters, for instance, in terms of limiting the number and types of targets it can engage. However, in case we make those modifications contingent on human intervention, our inaction might create humanitarian risks. On paragraph 7E, we have a question. We are wondering as to how this paragraph is different from the contents of paragraph 7C. In case there are no differences, we would suggest removing this paragraph. Also, we agree with the delegation of US that the reference to minimizing civilian harm does not apply only to this measure, but to all measures listed under this box, as do other IHL principles. So we request for more uniformity. Last but not the least, we have noted diverging views on paragraph six, in particular with reference to the term control and human judgment, on whether it works as an actionable obligation or as a conceptual framework or an idea. There are also varying views on its placement. We did make a comment yesterday on how this formulation could be unpacked and suggested its possible placement in box two. We can also base this reference to control and human judgment or the unpacked term that we suggested by highlighting the already agreed understanding that human machine interaction may take various forms and be implemented at various stages of the life cycle of a weapon. And that in determining the quality and extent of human machine interaction, a range of factors should be considered, including the operational context, the characteristics, and the capabilities of the weapon system as a whole. These two aspects, first the reference to control and human judgment, and the second to human machine interaction, the understanding that we agreed previously as guiding principle C, can work as a package formulation to elaborate control and human judgment. This idea was also proposed by the delegation of Japan, which we support. I thank you, Mr. Chair.

And let me thank the distinguished representative of India for sharing the views of his delegation, for making comments and proposals. Next on my list is Australia, to be followed by the US. But first, let me pass the floor to the distinguished representative of Australia. Madam, you have the floor.

Thank you, Chair. As foreshadowed, Australia would like to focus our second intervention on paragraphs 6 and 7, which concern context-appropriate human judgment and control. As we have stated previously, Australia supports the concept of context-appropriate human judgment and control as a pragmatic compromise. We understand the concept as referring to measures which may assist states to uphold compliance with IHL and not a distinct legal obligation. It is just one means for complying with IHL, but not the only means, nor is it a guarantee of compliance. We would therefore suggest some minor but high priority edits to reflect this. As we raised during the informal consultation and has already been referred to by a number of delegations today, We would propose that the words "to ensure" are replaced with "so that" and "are in compliance with" are replaced with "uphold compliance with." Paragraph six would therefore read, "Context appropriate human judgment and control is needed so that the use and effects of laws uphold compliance with international law, in particular international humanitarian law. including the principles and requirements of distinction, proportionality, and precautions in attack. On subparagraph 7, we particularly appreciate the changes that have been made, which now use language more familiar to militaries rather than unclear concepts that the GGE has been unable to reach an agreed meaning on in its debates. We suggest that 7 be corrected to say, "ensure the effects of laws can be adequately anticipated and limited." As we've raised before, effects may not always be able to be controlled due to intervening events. This would align the language here with paragraph two of box three, which similarly refers to anticipating and limiting effects. Further in respect of subparagraph 7B, we do not see a reason to include the wording number of engagements that laws can undertake autonomously. If it is retained, we suggest that controlling be inserted later in the paragraph, so that it would read, "by those responsible for their use, including by controlling, limiting, or otherwise restricting." This ensures that where appropriate, what is listed in that paragraph could be controlled, giving options for militaries to operationalize this paragraph. On subparagraph 7, we consider that the edits that have been made to be positive. However, to improve the grammar of the paragraph and ensure a technology-neutral approach, We propose that 7c would better read as follows: "Ensure that mission parameters of laws cannot be significantly modified by the system, including through real-time learning, without intervention by a human operator." This ensures we at least use the same language as box 1, paragraph 1, when we refer to intervention by a human operator. We also suggest that intervention may be more appropriately replaced with consideration or assessment, noting that a human operator may not always intervene if there is no need. And finally, on subparagraph 7, as others have raised, military objects is not a term used in IHL. IHL only refers to military objectives, and so we agree with suggestions by Singapore and other delegations to replace military objects with military objectives. Thank you.

I thank the distinguished representative of Australia for stating the positions of her delegation and making proposals. Next on my list is the U.S., followed by China, Germany, Singapore, and Ireland. But first, the distinguished representative of the U.S., you have the floor.

Thank you, Chair, for giving our delegation the floor. I know we've already had a fairly extensive intervention on Box 3, but we did want to respond to some of the comments that we've heard over the course of the afternoon, and in particular thought that we might start with some of the comments that we heard from the Russian Federation, which I think is a -- perhaps gives us a nice way to frame some of the points that we would like to make. So we heard a concern from the Russian Federation that the remarks of the United States suggested that the discussion of humanitarian control is not central to the debate that we're having. And I just wanted to clarify for the group that we do agree that this is central to the debate. It's not the only issue that's central to this debate. I think there are some key issues related to characterization, related to the identification and and the reflection of IHL principles. There are central issues related to the measures necessary to implement IHL. There are central issues related to accountability. But we understand that human control is a central issue to this debate. The second point that the Russian Federation made is that They are only hearing this now at this late stage of the debate. Really? Either we've been going to different meetings or we are not listening to one another. The United States has actively participated in this debate. I think we've probably participated in this debate more than most delegation would want us to participate in this debate. And we've been listening. To borrow a phrase that we heard at least six times in the Russian intervention on Box 3 on other issues, our position has been clear and has not changed. We would be more than happy to point any delegation to the consistent positions that we've expressed year after year after year from 2017 onward. Chair, we have listened. And what we've heard throughout the years and what we've heard again today only deepens our concern with this concept. One group, maybe a majority of the countries that are interested in this concept, many of whom initially focused on meaningful human control as the way to express the concept, have repeatedly explained that they view this as a principle of customary international law that's already emerged. And I think in that respect, it's very important to reflect on the remarks that we heard from the South African delegation this morning when not discussing this issue, to be clear, but when discussing the box one characterization points, I think was very eloquent in explaining that what we are doing here is not only debating what the nature of customary international law is, but what we do here will also be pointed to in the future as evidence of customary international law. And so one of the concerns that we've had is that for at least one stakeholder in this multi-stakeholder group, they will take any reference to this concept And they will describe it in later writings or in other fora as evidence of a customary international law crystallization, which we reject. We have persistently objected to this idea. Another group chair, and this is, I think, a view that we've heard emerge primarily among some of our friends from Europe, say that they see this as a principle that is derived from IHL. And I think we've explained why we don't agree with that. But nonetheless, that's another conceptual framework that some delegations have adopted in explaining this concept. Some delegations, and I think the Pakistan's intervention this afternoon, have explained it a little differently, that they see this as a very important new principle. And maybe it has some of its roots in some of these other theories that I've identified, but they think they're very clear. And we appreciate how clear they are on this point, that they see this as a new rule that should be included and should be agreed by states. Yet other delegations, like the delegation of Australia and its intervention just a few minutes ago, said that they don't see this as a legal rule, and they would only interpret this as a means to uphold IHL. Finally, the Russian delegation itself said, again today, that whatever this is, it remains at the discretion of the states parties to decide. That means it's whatever Russia decides it means. Chair, this is not a recipe for the progressive development of the law. Constructive ambiguity here, when we have at least these five differing positions and maybe more, is not a recipe to confirm a universal approach. It's a recipe for disagreement. It's not a recipe for unity. Thank you.

I thank the distinguished representative of the United States for stating the position of his delegation and reacting to views previously expressed by other delegations. Next on my list is China, to be followed by Germany. But let me first pass the floor to the distinguished representative of China. Madam, you have the floor.

Thank you, Mr. Chair. Our textual suggestions regarding box 3. On para 3, we suggest after "individual civilians" we add "protected by IHL". This would be more consistent with the provisions of IHL. As for paras 5 and 6, we understand we're not making any textual or a relocation of the text. However, we find para 5 and 6 are overlapping, we suggest merging them. As for para 7, we have concern about the formulation about the number of engagements. This position is not to ignore the risks brought about by the use of laws. It is based on the judgment of the norms of international law. From the perspective of international law, the key to the judgment of the legality of the use of force lies in whether the principles of distinction, proportionality and precaution are met. not on the number of engagements. If the use of a certain weapon runs counter to international humanitarian law, a single engagement should be prohibited. Therefore, restricting the number of engagements cannot address core humanitarian concerns And this would deviate from the centre of control and restriction. Therefore, we suggest deleting the number of engagements. Thank you, Mr. Chair.

I thank the distinguished representative of China for sharing the positions of the delegation and making proposals. Next on my list is Germany, to be followed by Singapore. But first, I pass the floor to the distinguished representative of Germany. You have the floor.

Thank you, Chair. Let me take this opportunity to thank you and your team and the friends of the Chair for your excellent work and leadership. Germany sees the current rolling text as being in well advanced stage. Therefore, we didn't comment on boxes one and two as we see them well formed. We did closely follow the discussion on the issues like the critical functions and lethality that still continue to be contentious to some delegations. However, we believe we are moving in the right direction. In Germany's view, box 3 is also moving in the right direction. We do have some comments on box 3, as well as some initial reactions to delegations' comments on this box. With regards to para 3, echoing also the comment made by Sweden earlier today, we would like to suggest to add, as in the other paragraphs, as required by IHL, as the protection of civilians and civilian objects, which is a paramount obligation of IHL, is not without exceptions which can be found in IHL. Para 3 would then read as, "It is prohibited, as required by IHL, to make the civilian population as such, as well as individual civilians or civilian objects, the object of an attack by loss." Regarding the addition in box 3, para 4, Like many delegations have argued before, we believe that the second sentence should be removed. We believe the scope and implications of this addition are not sufficiently clear. As stated during the informal sessions already, we would have rather seen any provisions regarding exceptions in the context of actual negotiations of an instrument rather than in the rolling text. It is a very complicated issue and discussion also containing questions about reporting, et cetera. So we do hope that this para will not eat up too much time in our ongoing discussions as we do have only very limited time left. Regarding box 3, para 6. Germany attaches great importance to ensuring that a framework of human control is in place during the life cycle of a weapon system. We continue to be of the view that there's a broad agreement within our group on the meaning of a human element in decisions about the use of force. The term context appropriate human judgment and control has emerged in the rolling text as a compromise in order to try and bring together different views on this concept. We, therefore, welcome the term being used in box three. However, we could also look favorably at suggestions to rewrite this paragraph while maintaining the notion of human judgment and control, such as suggested by Australia, if that moves us closer to finding consensus. We have taken note of the comments on the term context appropriate earlier. We understand this term as a suggestion to compromise. To us, context appropriate refers to the level and type of human involvement necessary considering the circumstances, such as the operational environment, the conditions, types of weapons, and mission planned. Echoing the comment by the Swiss delegation earlier, should rephrasing this concept, for example, by using the wording taking into account contextual factors as suggested by India earlier to bring us closer to consensus, we are willing to show flexibility like we've also done on other parts of the rolling text. Regarding box 3, para 7C, we could look favorably at the suggestion made by Austria to change human intervention to human authorization, following also the argument made by Norway and others who suggested that intervention might be solely understood as mechanical. We believe authorization to be more accurate, as for us, this implies an active element of approval. We are also looking into the suggestion made by Japan earlier today to rephrase that mission parameters of laws cannot be modified by the system autonomously. Furthermore, we could also look favorably at the suggestion made by delegations during the informals on para 7C to change significantly modified to substantially modified. And last but not least, we agree with the proposal made by Singapore, Japan and others today to replace military objects with military objectives in paragraph 7. Thank you.

I thank the distinguished representative of Germany for commenting on the text and making proposals. Next on my list is Singapore, to be followed by Ireland and the Russian Federation. But first, let me pass the floor to the distinguished representative of Singapore. You have the floor.

Thank you, Chair, for giving my delegation the floor once again. On paragraph 5, my delegation wishes to reiterate the suggestion we had made at previous sessions of the GGE for the word "may" to be replaced with "must." My delegation is of the view that the use of laws in accordance with IHL is not an option or a suggestion, but a legal requirement that states must adhere to. The paragraph could therefore read Laws must only be used in accordance with IHL, including the principles and requirements of distinction, proportionality, and precautions in attack. My delegation also listened with interest to fellow delegations' proposals on paragraph 7. We wish to react with a couple of comments. First, we would be favorably disposed to the proposal from the distinguished representative from Brazil to amend the phrase "mission parameters" to "target parameters." In our view, target parameters more accurately focuses our attention on the issue of concern, that is to say, ensuring that a LOS conducts its targeting function in accordance with IHL. In contrast, it is difficult to see how the more ancillary aspects of the mission would be of concern, for example, the exact path that the LOS adopts en route to engaging a target. Secondly, On the suggestion to amend the word "significantly", my delegation would be open to alternatives such as "substantially" instead. We would, however, be concerned if the term "significantly" were to be removed entirely. If every modification to the target or mission parameters of laws, no matter how minute and insignificant, were to require human intervention, This could compel belligerents to provide the broadest possible range of target or mission parameters with the intent of minimizing the need for human oversight. An example of an overly broad parameter could be target all military objectives. With such broad parameters, it could give rise to the effect that a LARS can autonomously switch its targets between a military aircraft, a military vessel or even a tank without the need for human intervention. Understood this way, a rigid requirement that human intervention is necessary for all and any modifications to the target or mission parameters could counterproductively result in less oversight. I thank you, Chair.

And I thank the distinguished representative of Singapore for sharing the views of the delegation, making proposals. Next on my list is Ireland, to be followed by the Russian Federation, Austria, and the UK. But let me first pass the floor to the distinguished representative of Ireland. You have the floor.

Thank you, Mr. Chair. We wanted to take the floor again under this box to go into a bit more depth on the debate over human control and respond to some of the suggestions made on the floor on paragraphs 7 and 8. On the human element, we want to take this opportunity to respond to concerns made over the last few sessions as well as today and yesterday, including, for example, that references to context-appropriate human judgment and control are legally unfounded, overly narrow or too vague to serve as a meaningful basis for our work. These counterpoints have been made well by other delegations, including Pakistan earlier. We want to take this opportunity to highlight them again, and so we request your indulgence. First, on the argument that IHL does not explicitly require human control as such, we agree that those precise words do not appear in treaty text, but that is not the relevant question here. IHL regulates human conduct in armed conflict. Every core obligation, distinction, proportionality, precautions in attack, presupposes a human agent capable of making legal assessments and being held responsible for them. The issue before us is whether increasing autonomy risks weakening the chain between human legal obligation and battlefield outcome. If autonomy risks interrupting that chain, then articulating the need for context appropriate human judgment and control is not inventing a new rule. It is safeguarding the conditions necessary for existing rules to function effectively. Second, we do not agree that the concept is too vague or would lead to divergences in implementation. The rolling text itself clarifies and operationalizes the concept. Box 3 does not leave the concept undefined. It elaborates the practical elements required for it to exist, including requirements related to understanding system capabilities, temporal and spatial limitations, setting operational constraints, and enabling effective human oversight and intervention. In other words, the concept does not float abstractly above the text. It is given concrete meaning by the measures that follow it. Removing the concept while retaining those measures would strip the text of its organizing logic. Conversely, keeping the concept together with its elaboration ensures coherence and clarity. Third, on the concern that the formulation is overly narrow, context of proper human judgment and control does not exclude software execution. It does not require continuous manual operation. It does not imply that a human must physically direct every technical function of a system. Many lawful weapon systems already operate through automated or software-driven processes, and that has never been in question. The issue is not whether software may execute a decision. The issue is whether humans retain meaningful responsibility for the legally relevant judgments that precede and shape that execution. If software operates within parameters defined by a human, if the human understands the system's capabilities and limitations, sets operational constraints, determines the target, and remains accountable for the outcome, then software execution is entirely compatible with context appropriate human judgment and control. And that is precisely why existing systems, including guided munitions, are lawful. However, if a system substitutes for human legal judgment in selecting targets in complex and evolving environments, if it determines which specific object or person to attack without a sufficiently informed and constraining human role, then the structure required for IHL compliance changes in important ways. IHL attaches obligations to humans. Software may assist in implementing those obligations, but it cannot replace the human responsibility that IHL presupposes. For that reason, context, propriety, human judgment, and control is not a narrow or competing standard. It's a structural condition that ensures IHL obligations remain meaningful as autonomy increases. On the point that autonomy may increase protection of civilians and to avoid privileging human involvement over civilian protection, For us, context-appropriate human judgment and control does not reject autonomy. If autonomous functions can enhance precision and reduce civilian harm, that is, of course, welcome. But improved performance does not eliminate the need for accountable human legal judgment. Precision without accountability is not sufficient under IHL. And operationalizing context-appropriate human judgment and control through the measures in box three helps remedy that gap. We also agree that human judgment and control are not ends in themselves. They are structural means through which IHL obligations are implemented. For us, human judgment concerns the legal assessment required under IHL, and human control concerns the practical capacity to ensure that system behavior reflects the assessment. For these reasons, we strongly support retaining the concept of context-appropriate human judgment and control in the ruling text. together with the detailed measures that give it substance and operationalize the two-tier approach. And we reiterate the point made earlier by the ICRC that this formulation is a product of significant compromise, and returning to previously agreed language here is insufficient to meet the challenges posed by loss. On the specific formulation in paragraph 6, we take note of Australia's suggestion, and we'll look at it carefully. And we reiterate our objective is to ensure that existing obligations remain effective in the face of emerging technologies, which is indeed one of the main purposes of the CCW. So just a few reactions on paragraph 7 and 8. Paragraph 7A, we wish to retain the reference to ethical. We can support Austria here and as well consider Singapore solution. 7C, we agree with others to remove the word significantly, or otherwise we can use the word substantially. And we also agree with Austria and Norway that human authorization might solve the issue of nominal human input here. We also wish to keep the paragraph 8 on design and development. Compliance in use cannot be divorced from choices made in design and development, particularly from systems whose behavior is shaped by software. And we do think that the Australian suggestion can be helpful here in combining this paragraph with paragraph 2 from box 4. And I'll leave it there. Thank you very much, Chair.

I thank the distinguished representative of Ireland for sharing the views of his delegation and also responding to previous remarks. Next on my list is the Russian Federation, to be followed by Austria. Let me first pass the floor to the distinguished representative of the Russian Federation.

Thank you, distinguished Chair. We'd like to respond, clarify some of the comments made by the delegation of the US. We think giving these explanations is very important because those explanations concern certain misgivings expressed by our delegation and as confirmed in the statements by a number of other delegations before us. Such misgivings are felt, it seems, not just by us, but possibly by significant number of delegations actually sitting in this chamber. The that are possibly misgivings that are possibly felt by a large number of other delegations sitting in this chamber. First of all. Our misgiving. Don't concern the fact that the US. supposedly is calling into question the central nature of the human element for discussions within this group. We didn't say that. Maybe something was lost in the translation or perhaps the theme of the human element are sometimes complex to understand. So we'd like to repeat Our misgivings concern the fact that, it seems to us, the US delegation is calling into question the human element in the context of ensuring compliance with international humanitarian law. For us, it is very important to understand to what extent they are calling into question the fundamental nature of the human element and of human judgment for the purpose of ensuring compliance with the principles of international humanitarian law. It is important because We already have an agreed consensual understanding within this group, which clearly sets out that human judgment and human machine interaction are the fundamental essential for ensuring compliance with the norms and principles of IHL when activating such weapon systems or in the context of such weapon systems. If this is being called into question, that would be a very serious factor that would change the situation within this group, if that is the case. It is also important to understand that, in the context of the document that you have presented, the human factor, human judgment, is one of the fundamental elements. The human factor effectively lies at the foundation of many of the provisions in your document. What is more, once again, human-machine interaction is one of the essential fundamental elements of the guiding principles that were agreed by the States parties to the convention in 2019. If all of that is going to be called into question, If we are saying that that is no longer relevant, if the concept in itself of human judgment, human-machine interaction and the human factor in the context of ensuring compliance with norms and principles of international humanitarian law are being called into question by the delegation of the US, and the last comment only appear to convince us further of that, since it was said that the US delegation has concerns or misgivings about that concept. As we said earlier, we have serious concerns that we are not addressing what we should be discussing. where we are in the discussions in this group. Because statements like those, if we are to understand them correctly, that would seem to completely void all the previous work done in the group and would therefore create a completely different playing field in our discussions. Our second concern concerns the overall understanding that was reached by consensus in the group and that was enshrined both in its reports and in its guiding principles that were adopted by the States Parties to the Convention. Our misgivings are that, it seems to us, such understandings are also being called into question. If we are mistaken, please confirm to us one way or another. We would be very happy if we are mistaken. However, if we are not mistaken, we would like to understand that right now, because that would be a very serious factor having a direct, fundamental influence and impact on the work of the group. Our third misgiving concerned the fact that that was being done specifically at this stage of the group's work, when we find ourselves just one step away from agreeing on starting work on the draft document of the group. misgiving is that if everything we said earlier is reality, then we simply don't have time to rectify all of it. And we are seriously concerned about all those three elements that we have just highlighted, because that will have far-reaching consequences and will require almost a complete upheaval in the future work of the group. Are we mistaken in what we have just said? As we said, we hope we'd like to be mistaken, but if that's not the case, then we do want to understand where we stand, and let us face the bitter truth in the eyes. Thank you, Chairman.

Let me thank the distinguished representative of the Russian Federation for responding to previous interventions of delegations. We are a bit at a moment of a cliffhanger because this is literally, this is, but this is not two for 12, this is two for six. So I think that's the good news. The other good news is that you will not be provided with food and drinks, but you will be provided with food for thought as of seven o'clock. As of 7 o'clock in room 11, and this is for high contracting parties and the International Committee of the Red Cross, we will turn into an informal mode. Some of you have already requested that. We had planned for it already because this is, I think, really the moment to start doing so. We are not continuing the discussion we had until a few moments ago. We will go back to box one, and more in particular to para 1. My team and I have conducted consultations in the corridors, and we have come up with a text that we want to test with you. So please feel free to join us I'm fully aware that we still have at least two more requests for the floor, and there might even be more after a good night of rest. So we will continue our discussion with regard to box 3 tomorrow morning at 10 o'clock. I will now adjourn the meeting. Yes, I have one very practical one. If you leave the compound, make sure you re-enter before 7 o'clock or you will have to climb the fence. Because I think the gate closes because of austerity measures, et cetera, et cetera. So please bear that in mind. That was, I think, all the technical things we have. Yeah. So then I join the meeting. See you in room 11 later on, and if not, tomorrow morning at 10 o'clock in this room. The meeting stands adjourned.