5th Meeting - 1st Session Group of Governmental Experts on Lethal Autonomous Weapons Systems 2026

Excellencies, distinguished colleagues, good morning. Could I ask you to take your seats again, then we can start. Good. Good morning again, excellencies, distinguished delegates. I hope you had a good night's rest. Thank you. Many of you made yourself available yesterday evening where we had some informal discussions regarding Box 1 and particularly regarding subpara 1. My intent is to give a short feedback on that after we have completed our work on Box 3. It's not my intention then to continue the discussion on that, but it's just to give you a feedback because the discussion we will then continue tomorrow when we have second reading. And then after that, start with the reading of the boxes four and five, all with the intent to finish the complete reading of the text, box by box, today before six o'clock. Then, one more technical thing. If you have prepared statements, the interpreters would highly appreciate it to receive them in advance. I think it would also contribute to the quality of the interpretation and facilitate the work of the interpreters. So, if you have something precooked in your tablets, please join it with the interpreters. And then I would now like to... And one other remark, please remind that they need time to translate. So don't speak too fast, because some of you, I must say, if it's then translated, it's not always easy to follow. So also for the impact of your statements, I think it's helpful if you bear that in mind. And then I'm looking at the list. I have three more requests for the floor from Austria, United Kingdom and the United States to continue our discussion on box number three. So let me now first pass the floor to the distinguished delegate of Austria. You have the floor.

Thank you also for all the efforts and last night consultations which have been very.

Rich and intense and also quite.

Enlightening and I think we have made some progress at least on box number one. Coming now back to our discussion on box three, which is definitely the most important discussion that we are having, which was also quite intense yesterday with some strong views and many new proposals. I think the intensity is also a sign that we are getting closer to a result, even though it might not look like this at the very moment.

But I think this phase is also necessary until we come to a.

Very good result at the very end and make the result in the end also even better.

We just want to react to the discussion and also the proposals that we heard.

There have been some detailed proposals also by the US delegation. Many of them are, in our view, useful when it comes to resolving, um, issues, issues that we are carrying with us for quite some time, um, and the text that they have provided us with, uh, is definitely something that we will look into. and where we can also look for solutions for problems that we have carried for a long time. In other cases where we did not have carried intense problems for a long time, we might rather stick to the language with which we have worked now for some months and years. And just maybe going into the first three paras where we had not that many discussions or rather discussions so far, we're rather moving different elements around. One remark that we wanted to make when it comes to those is that we hear all the comments of changing them more closely to Article 51 of Additional Protocol I and into existing IHL. I just want to recall that the original idea, I think, from you and your team was that this was rather more closely aligned to paras, which we have already in Amended Protocol II or in other parts of the CCW, which to some extent is also what we have in our mandate. And the idea there, and we have made this point already before, is that we are specifying, we are going deeper on those key principles of the CCW, where all the other protocols are looking into the unnecessary suffering and indiscriminate effects of weapon systems. And this is definitely something which can be deepened and specified, and we heard some also some good proposals by Pakistan and others in this effect. Then going into some of the proposals that we heard, or maybe the spirit of these proposals. The US proposals are in many ways very correct. and are reflecting IHL as it exists. And I think this is the core of our discussions. To some extent, they are also very clinical. They are very clinical and clean insofar as they are reflecting the IHL as you know it at this point. But as we and others in Pakistan and also Ireland made some very good points yesterday, what we are trying to do here and what the CCW was designed to do was to go further than that. And this is not only since there is a need from a humanitarian side or also from a human dignity side, I mean, something that has driven this process since the very beginning, but is not that present at the moment, but definitely is a driving force. There is also, as Ireland has very much, very well spelled out yesterday, there is the issue of preserving IHL, making it applicable to these new technological developments and solving a problem that has not existed before, but is now in front of us due to technological progress and the issue that the law as we know it was designed for humans and not for machines, and we just need some translating effect into this. Some general points here again, but I think they're important because they address a little bit the spirit of the amendments, which are very logical, but then are not ambitious enough to address these kind of issues. and I also reread the draft articles which were presented by a group of delegations in previous years. To some extent, we feel that they have more captured this idea and these problems and how to resolve them than the current proposals, and we might refer to some of these proposals later on in the text. Other main point, and why we need to do this, uh, and why it is re- uh, necessary, uh, to look more deeply into, uh, new issues and also into, uh, elements that are not currently spelled out that, uh, specifically in IHL is what is in, uh, para 8. Um, we heard some, Uh, proposals to move this to box two, uh, we would not agree, um, some of the proposals by the US to make maybe a little bit more specified which prohibitions and requirements are meant here is maybe useful, but in general, this is a key part of box three to us. We also would have liked it more prominent in the beginning of box 3. We could agree with having it here at the very end as a catch-all, but it's definitely a box 3 issue and not box 2. The reason is the one that I have spelled out already. We are facing new problems. We are having a different kind of technology. Very many of the issues which are related to IHL are already predetermined or fixed in at the development and design phase. And here I also have to point to the draft articles document, where it's in article one and article two has design and development as a key issue that has to be resolved by our work here. So as it is a key issue, it should stay in box three and not in box two. Next issue is the discussion about context appropriate human judgment and control. This issue also belongs into box three. I think there's not much leeway for us. We would have liked it in box two and in box three. We had a lot of language on this, which was much more stricter. And what we have in front of us is the last compromise that we can take on this issue. There have been some comments also that delegations have five different positions on what human control means and how it applies. Our delegation also had already some different positions on what this concept means and what it is in a legal way, as it has spelled out by the US delegation. But the reason why we are moving through different phases on how it applies and how it is needed is that we are trying to seek consensus and compromise here. For us, it would have been appropriate to make this a prohibition part, since it is so crucial to have IHL still relevant in the day and age of artificial intelligence. That we are now have moved into it as a principle and a concept that translates it or even less. I think this is a part of our discussion and a result of our discussion and also a result of us compromising. So this should be appreciated and not criticized, I have to say. That it is a concept that is relevant. I mean, we, I think what is more important is to look into the basic idea. In one of our previous working papers, we have spelled it out in such a way that the effects of the use of force need to be traced back to human agency. We don't need this text here. I mean, it's not something that is useful for this text, but the idea should be that there needs to be a connection between a decision by a human and the use of force. And if we, this kind of connects, this nexus, this connection needs to be made somewhere in this text. Most of us agree that context appropriate human judgment and control is doing this and is the right and the most flexible concept at this very moment. We can look into other other ways, I mean, we very much would appreciate keeping this concept and not moving too far astray from it. I also need to point out that the idea that I had just spelled out to some extent is also reflected in the draft articles where it talks about humans and operators, operators and commanders, and they're also trying to establish this link. The step from talking about operators and commanders going to talking about humans is definitely one that is probably difficult for some delegations, but it's one that we have to make. And I would just plea to make this step in our work here and as soon as possible. Sorry for being so long, but I have a few other points.

We also proposed to.

Take the issue of the ethical consideration, which is currently under box A, and make it more prominent. Here, we want to point out that this has been part of our work for a very long time. I mean, currently, it is more an afterthought. And I also want to recall that high contracting parties during the 2021 review conference have also solemnly declared that international law, in particular the United Nations Charter and international humanitarian law, as well as relevant perspectives should guide the continued work of the group. So we all have solemnly declared that this is an element that we should look into. So we just recall that and would plead to keep it in the text and have it also continuously as a part of our work. What we hear, we hear some proposals about good faith and due diligence. We still have a question mark on due diligence. But the good faith element could be useful insofar as we have always requested to also have the element in under seven and eight under the points that follow that goes in direction that there should be an understanding of the system of the intended effects and so on. So maybe this could be something that could be introduced since others have proposed it as well. But it's definitely new text that has to be looked into when it comes to the discussion around the number of engagements. I mean, here we also heard some good proposals by the US that keeps the text in their amendments to B. We still think that it's extremely necessary to have such a limitation. And just to explain it once more, the issue here is that B sets certain parameters under which humans can still exert some kind of judgment and control. The reason is that the specific specificities of this technology and algorithm can process within seconds many, many different decisions or even engagements. So by limiting them, you're setting a safeguard. And currently we are not saying what kind of safeguard and how high the safeguard needs to be. You can, as I pointed out before, you can set it at 1000, 2000, 5000 or 1 million. The simple step that we think is necessary here is that whoever operates the system has to make this conscious decision in this context, in this situation, I don't want to have unlimited engagement because in that situation I cannot anticipate and control what is going to happen. So if I say this kind of weapon system, and this is again the context appropriate, will not pose this problem, so I'm not necessarily have to limit the number of engagement. If it is a system with a high impact, which has unlimited, even energy or which is not limited by ammunition or anything else, I can make this limit and set it at a certain scope. On C, there has been some discussion on the mission parameters and there were some proposals to make it more specific on on on on targeting just to point out that the mission parameters that the systems should not change should not change could also refer to a geographical scope. So it means this is the mission domain in which the system is employed and if a system then changes the mission parameters which also means the geographical area by itself, you could actually get into some trouble when it suddenly moves itself into a civilian densely populated area or something else like this. Also, we want to support the very good points made by Belgium on real-time machine learning. And I think there's still some misunderstanding what real-time means in this context and also how machine learning works. What we are not saying is that there should be no machine learning and the data collected on the battlefield should not be used. There is the possibility via machine learning, and it's not only a possibility, it's probably also a very strong recommendation that the data that you collect during active operation is not fed into the system right away, but at the later stage. That means this data, this experience that the system has is not lost, it just is not fed into the system right away, but in a more safe environment. And this is highly recommended as many other elements here under seven. as measures that are very reasonable and that many militaries will implement anyway because it's the right thing to do. You don't want to have a system that is unpredictable and just goes off and changes its own modus operandi, which could even lead to friendly fire and other problems.

This was our last point.

Maybe just following up on the context, appropriate human judgment and control and what Eilidh has said. where we derive this from. I think one element that does not deserve enough, has not enough attention, why control is something that we can derive from existing IHL is also what we find in Additional Protocol I when it comes to the predictability of weapon systems and the intended targets. I just don't have the text in front of me, but there are some formulations in there which very clearly set out that you should be able to target a weapon systems to a specific target. And I think this is one of the key elements here that we are talking about unpredictability and that you're employing a system where you not necessarily know when and how and what target it will strike. And this is definitely something that is not in line with IHL and something we need to address. through maybe even, uh, a new concept. Um, sorry for being so long, maybe I will be more specific on the last argument later on since I had not, uh, had the text in front of me, but I think it's, uh, definitely something that is important. Thank you.

Let me thank the distinguished delegate from Austria for reacting to previous statements and, uh, making proposals. Next on my list is the UK, followed by the United States and Sierra Leone, but let me first pass the floor to the distinguished representative of the UK. You have the floor.

Thank you, Mr. Chair. We haven't intervened up to now because there was much for us to reflect on in the interventions of yesterday afternoon. In fact, even more for us to reflect on in the intervention we have just heard from our Austrian colleague as well, so I hope you will forgive me for taking a slightly reflective intervention. We would like to offer some observations of what we've heard in relation to box 3. The UK sees box 3 as important in grounding this discussion in the rules and principles of IHL. The reiteration of IHL here frames the measures that follow. with the concrete prohibitions and the necessary preconditions that govern the conduct of hostilities and the use of laws in the planning and prosecution of hostilities. That box sets the foundations of the set of measures. We recognise that getting the balance right means understanding the law, understanding what is critical to comply with the law and understanding the malleable criteria that support the effective and lawful use of capabilities but which depend upon the circumstances. Herein, we have the necessity of the understanding of context. It is the point at which those of us who are lawyers will say, "Well, it depends." It depends on the circumstances, it depends on what you're trying to achieve, it depends on what system you intend to deploy and how it works. First, to make some points on the earlier paragraphs in the box. On paragraph three, we note the comments made by the US that this paragraph could be grounded in the principle of distinction. We endorse that and we endorse Australia's suggestion of adding "protected as such" after individual civilians, which better articulates the principle and allows for the potential loss of protection. We do not support suggestions that paragraph 3 or anything within that box should include further text in respect of the targeting of persons. Not only is that not a reflection of the existing law, which is our understanding of the objective of the paragraph, but the group is some way off consensus on mandating such a restriction. I move on to box 4 and the text on retention, acquisition or transfer permissions for the purpose of training and development of countermeasures. Like other delegations, we understand that this type of language reflects similar provisions in other legal instruments, such as the Ottawa and Oslo conventions. But also, like other speakers, we are conscious that this concept is strained when the relevant component of the capability that you're talking about is amorphous, as it is with technology, rather than a product with specified physical and technical characteristics. At the point of determination that a capability cannot be used in compliance with IHL, certain questions need to be asked. Is the capability simply so unpredictable that there is no case in which it could be used with discrimination? If that is the case, it is not clear to the UK what value there is in maintaining the capability for the purposes of training or countermeasure. both of which would need to be predicated on the idea that something in the behaviour can be predicted or planned for. Separately, and of particular relevance to whether this concept can be meaningful for laws, the stage at which behaviour of a capability is determined for autonomous weapons is not the same as the stage at which behaviour is determined for other weapons. Those issues are reliant on training. on the proper processing of sensor data, on the match between those issues and the parameters set for any particular operation. When applying rules of, for example, Article 57 of Additional Protocol I, autonomous weapons systems can be thought of as empty vessels waiting to be populated with behaviors, the commander's intent, and the legal parameters necessary in the circumstances. The item that might be retained or transferred is largely this empty vessel. This feature of these systems renders these measures redundant. The UK recognize that the aims of many states in this room include addressing concerns about proliferation and is open in the spirit of compromise on this provision. However, we would suggest that whatever is retained in this paragraph could be reduced in length and simply refer to manufacture, acquisition, stockpiling and transfer of systems that cannot be used in compliance with IHL, and it could be moved to the subsequent to paragraph 5 to sit more naturally with the flow of this section. So that brings me to paragraph 6. The UK believes that the removal of context-appropriate human judgment and control from the second box improves the structure of the document generally and better reflects what we understand to be the intent of the phraseology. We agree that the concept should not be held up as a legal rule or a fundamental IHL principle in itself. It was encouraging to hear from various delegations yesterday to suggest that that was not, in fact, the intent that they had attributed to the statement. The concept cannot, therefore, be held up to be an end in itself. It is more true to the discussions of the group to set out measures which are required for the achievement of the goal of compliance with IHL, or which may be helpful for achieving lawfulness, or even further policy standards. In our intervention on Monday, we spoke of the need for a bridge from the law to that concept. The UK recognises the context-appropriate formulation, because it is similar to the concept of context-appropriate human involvement, which is central to our national policy approach. The UK has set out what we mean by our concept. It focuses on the role of the human as the accountable actor and the effector of the obligations that sit with the state under IHL. It therefore frames the necessity of human interventions to the servicing of their legal obligations. Guiding principle B and the draft articles that were contained in the joint paper provided by us, the USA, the Republic of Korea, Canada, Australia, Japan, Estonia, Latvia, Lithuania, Poland and Czechia, each recognise that human responsibility for decisions on the use of weapons systems must be retained, since accountability cannot be transferred to machines. We recognise the arguments of colleagues from Pakistan and Ireland that these systems raise questions in regard to the legal position which we previously thought to be inherent. The UK also recognises that the reason that this group of experts was formed was because autonomous weapons systems act in a way that appear to substitute the role of the human with the functioning of the machine. We know that this is, in fact, not the case, insofar as the human's role as the legal agent simply cannot be transferred to the machine. We acknowledge the US argument that the presentation of this concept as an end in itself presents it as a standalone rule comprised of ephemeral standards such as appropriateness. We acknowledge that the term context in this formulation is different to the term context used in the consensus language that we have found in this group, including in guiding principle C, which has referred to operational context among the factors that should be considered. In determining the quality and extent of human-machine interaction, the other factors are the characteristics and capabilities of the weapons system itself. The UK does not believe that it is the intent behind the concept of context-appropriate human judgement and control to similarly limit the term context to operational context. Rather, our understanding of context here is that it includes all factors that are relevant to the servicing of the end of lawfulness against the inevitable variety of factors that will arise from different systems, overall objectives and operational contexts. Therefore, while the UK accepts and can accept the language of context-appropriate human judgment and control in service to the achievement of consensus and the need to capture the core issue that has exercised this group under this and previous mandates, we believe that we need to find a mechanism to provide the grounding of the concept in the effect that it is seeking to achieve. We need to ensure that we have a sturdy bridge to the legal ends. We therefore reiterate the point that we made on Monday in box 2 that the state's obligation to ensure respect is affected through individual accountability, including through a responsible chain of command and control. We need this to set the context of the concept. To adopt this approach then makes paragraph 7A slightly repetitive. Indeed, the other measures contained in paragraph 7 are specific to the manner in which the autonomous weapons system is controlled, as we heard yesterday. They reflect the sort of feasible precautions that we might apply. 7 is more fundamental than that, however. Therefore, if we take the suggestion of elevating an amended version of 7 in paragraph 3 to box 2, It does not need to be repeated here. On 7, the UK agrees that the language about the anticipation and control can be clarified and simplified, and therefore suggests that the paragraph is amended to read: "ensuring those responsible for the use of laws can control their effects, including 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 laws can undertake autonomously. We will provide that language to you, Mr. Chair. On 7C, We agree that the term "significantly modified" is unclear and does not conform to the standards that might otherwise be applied in areas such as legal review, which would require material modification. We have listened to our colleagues, in particular those from Brazil and Singapore, including on the need for a modifier. However, we believe the sentence can be further simplified by simply focusing on the critical functions of the autonomous system. Therefore, I suggest that the paragraph might be amended to read: "ensure the mission parameters of laws as pertaining to targeting, selection and engagement cannot be modified by the system without human intervention". Finally, on this box, we also agree with the suggestion in 7e to replace "military objects" with "military objectives". Thank you for indulging my long intervention, Mr. Chair. We will send our suggestions in writing.

Let me thank the distinguished delegate of the United Kingdom, and indeed, we would appreciate to receive your thorough thinking in writing. Next on my list is the U.S., to be followed by Chad, the EAU, and Sri Lanka. But let me first pass the floor to the distinguished representative of the U.S. You have the floor, sir.

Thank you, Chair, and good morning, colleagues. So I know this is the third time that we've taken the floor to intervene in this discussion, so we'll try to be brief. But our delegation did want to respond to several of the questions that were posed by the Russian Federation at the end of the evening last night. And then listening to the debate this morning, I think we wanted to make a few additional observations. So first, Chair, just to recap, at least the way we heard the Russian intervention last night, which was framed, I think, primarily as a request for some clarifications from our delegation. The way we heard the questions expressed, I think they were expressed as misgivings, was first, were we moving away and rejecting the specific conclusions that this group has reached in the past regarding the human dimension? Second, again, expressed as a misgiving, were U.S. statements intended to convey to this group that we were rejecting all of the previous understandings and decisions of this group? And third, He said that if they were correct about these misgivings, that it was time for this group to stare the bitter truth in the eyes and to ask whether we were at a catastrophic moment in our discussions where the position of the United States was risking the outcome of our process. And so, Chair, maybe just to take these in reverse order. With respect to the last statement about where we are in the process, we were a little surprised to hear a concern like this on the second day of 10 days of meetings in 2026 at the beginning of the third year of our negotiations. From our perspective, we understand there's time pressure and believe us Chair, we feel the time pressure, but I don't think that we're quite at the moment of catastrophe. I think we have time to work through these issues, and the practice that we've observed in this group over the years is that we typically use all of the time that's available to us to work through the issues. For all of us who have been in the late night sessions on the last day of a particular year, a particular process, I think we all remember that there are often issues that are very difficult to resolve, that may not be resolved until later in the process. And we hope to avoid that. I think we hope to make progress earlier. But it seems a little bit premature to this delegation to say that we have reached that point on the second day of our negotiations, or that we have reached that point because of the attitude of the United States towards these negotiations. Second, Chair, on the second more general misgiving, whether the interventions of the United States should be intended to call into question the commitment to the prior decisions and the prior agreement that's been reached in this body. That one to us sounded, didn't correspond to the reality that we think we've seen. I think yesterday I mentioned that I wasn't sure if we were going to different meetings or if we just weren't listening to one another. At this point, I'm not sure we're even in the same reality. The United States has been one of the clearest voices in this room about the importance of the past decisions and past agreement in this -- in this group. We are so focused on these past agreements that we have tried very hard to build them in to this rolling text. I think what has surprised us, Chair, is the reluctance of this group to use the foundation that we've built in the past to anchor this document. And so many of the interventions we've made have been to make that point. And I take the point that the Russian Federation made yesterday that, well, there's a simple way to do this. We simply have something at the very end that recalls all of the previous decisions and all of the previous agreed principles. But I think our point is a little bit different. And it's the first misgiving that I think really illustrates the point that we've been trying to make. So the misgiving that we were rejecting the agreed principles related to the human dimension. Chair This was exactly what we tried to do with our first intervention on Monday, where our recommendation to address the human dimension was to include the exact language that the Russian Federation quoted yesterday. So far from rejecting those paragraphs, those were the paragraphs that we were focused on. Those were the paragraphs that for us were important to reflect in the document. So I hope this gives the Russian delegation and the rest of the room some assurance that we are not walking away from previously agreed principles, either related to the human dimension or more broadly. I also wanted to react to a few of the points that were made by our colleague from Austria. And the first reaction that I wanted to share was really just to express our appreciation to him for his continued engagement on these ideas. I think the Austrian delegation has been one of the most serious and most focused delegations, and we appreciate the ability to really engage on the issues. And we don't always agree. And in fact, I will probably describe a few areas right now where we don't agree. But it's precisely this back and forth that is what makes this body so valuable. And it's precisely this back and forth that is ultimately going to get us to the consensus that we need to move forward. So we did want to thank the Austrian delegation for that level of engagement. And I do want to make very clear that to the extent that we're pointing out what we see to be differences in the approach to human control that we see reflected, I identified five different perspectives yesterday. There are probably more. We heard another one this morning from the UK delegation. It is not a reflection that we don't appreciate the need to engage on these issues. It's not an expression of the that that we do not appreciate the positions of other delegations. We are not criticizing anyone in this room for any position they're taking. So please, we are not criticizing anyone. We are grappling with difficult issues when we have a position We are going to say what we think is right. That's all we're doing. If we disagree with you, it's because we have a principled difference of opinion. We are not being critical. I'm reminded that Pakistan has very, very helpfully focused us on the three categories that really are at the heart of the current rolling text. We have organized a lot of our internal thinking around those three categories because we think it's, and we believe it's reflected in the rolling text. Our Pakistani colleagues have identified the need to, in a first category, identify existing international humanitarian law. In a second category, identify the measures that should be taken by states to achieve compliance with existing IHL. And in a third category, they've pointed out that where those rules and those measures fall short, there is a need for new legal rules. We understand their position. We respect their position. But where we have respectfully disagreed is on whether We have established in this body that existing IHL is insufficient, and there is a need for the third category. And indeed, the debate that we're having about human control illustrates the importance of the framework that Pakistan has presented to us, because we have some delegations in the room who believe meaningful human control is already an existing IHL principle. We have some in this room who think that it's not a legal principle, but should be a measure designed to achieve compliance with IHL. We have still others in the room who believe that it's maybe neither of those two things, but it certainly should be a new rule. We just disagree. Chair, I think the last thing I wanted to mention is the concern that has been raised for the clinical and the technical approach that we've taken to these discussions. Chair, we are talking about legal rules. From our delegation's perspective, not only is there nothing wrong with being clinical and technical, when we are identifying the contours of legal rules, we think it's our responsibility to do that. And to the extent that we are seeing things that we don't think are correct, we are going to keep making those points. But we're not the only ones making those points. Everyone in the room intervenes and makes these points. I don't know why when we do it, it's viewed as a delegation being clinical and technical. But when others do it, it seems not to pose a problem. So we would just simply ask for the same for recognition that we are doing something that is similar to what everyone else is doing. So, Chair, I'll stop there and thank you for giving us the floor.

Let me thank the distinguished representative of the United States for sharing the position of his delegation, also responding to remarks yesterday. I think more in general about what we are doing here, I wrote it down for myself as that we are clearly not only restating existing IHL rules, but also clarifying how they apply in the context of laws and to articulate specific measures to respond to the particular challenges that laws raise. And I think that is a very challenging mission, and we have tried to reflect that. And indeed, I think our Pakistani colleague referred yesterday to, I think what he said already during our first session during this mandate in 2024, he called it then three boxes. And indeed, I also recognize that there is a difference of opinion with regard to that third box. But I think for the time being, we have a lot at hand to satisfy, I think, the kind of a mission statement I have for myself and I just phrased. And yes, we still have eight more days, but we also still have a lot of work to do. And I operate in a different way than my Brazilian colleague in the BWC working group, where he's working with colleagues. Some of you already smile when I say that, because then I don't know how much green we have. And we have we are visiting and revisiting certain texts. And then we still have certain discussions. So there comes a time when we have to look in, indeed, to what we have on the table. And indeed, what is agreeable. to bring forward. And there I have also earlier stated what is my level of ambition, indeed, that we, with this mandate of three years, we have something to show for. Yes, a lot of work was done in the past, and we might find a way of also recognizing that in our report. but we can't just refer to that or just reproduce it. That is, I think, not our mandate. Our mandate is clear, a set of elements of an instrument without prejudice to its nature and other measures. And that is indeed not an easy job, especially not in the present geopolitical situations. But I think we have a text at table at this moment, where there is enough substance in it to achieve that. And of course, we have to take into account the positions of all delegations, because that is also one of the things that is in our mandate, it's consensus. And I keep saying that consensus doesn't mean that everybody has to love everything in it. Consensus, in my definition, is the absence of dissent. So in the end, on Friday, in September, will you be able not to dissent? That is what we have to aim for. That is what we have to work on. And I think that I really appreciate the discussion that we also have on what to achieve and what is the backgrounds from the different delegations. But then I, I would really like to ask you again to focus again now on the text. Uh, I think we had some interesting remarks also, um, from the perspective of the UK delegation this morning and the Austrian delegation. Um, I think, uh, bear that also in mind, uh, as the US colleague said, we have five, possibly six, um, um, um, uh, proposals. approaches, perspectives. Let's try to construct out of that something that is palatable to all of us. And with that in mind, let me continue with a list of speakers, of a list of interventions, Sierra Leone, Sri Lanka, and Bangladesh. But let me first pass the floor to the distinguished representative of Sierra Leone. You have the floor.

Thank you, Chair. Sierra Leone views Box 3 as a cornerstone for the rolling text. We welcome the retention of key prohibitions in the two-tier structure, combination prohibitions with positive obligation to ensure meaningful human control. We would like to add our voice to the ongoing constructive conversation on this box. On paragraphs 1, 2, 3, and 5, we support these prohibitions and their retention from earlier versions. These provisions restate the fundamental IHL principles we believe applicable to laws and should be preserved. On paragraph 2, Sierra Leone particularly welcomes the retention of this provision from the May text. This prohibition using language from Additional Protocol I is crucial for prohibiting unpredictable autonomous open systems that prevent users from understanding, predicting, and explaining system outputs. However, in our view, this is there is a critical gap regarding the removal of explicit, explicit language on ensuring systems are adequately predictable, reliable, traceable and explainable to those responsible for their use. These characteristics, which reflects consensus among many states, were recognized in the 2023 GGE report, as mentioned by several delegations. and we believe are fundamental prerequisites for meaningful human control. We therefore support the proposal by delegations to reinsert this language to the rolling text. On paragraph four, retention, we share the perspective of delegations who spoke before us that unlike conventional weapons such as cluster munitions and alike, autonomous weapon systems are not countable items, but often consist of software-driven systems, integrated components, and technologies that are continuously developed and updated. Concepts such as minimum quantity and destruction may not be applicable, and the ease with which autonomous systems and related capabilities may proliferate suggests that a retention framework based on limiting quantities or physical destruction may be less effective. We therefore support the proposal from delegations to remove this sentence. Additionally, on anti-personal autonomous weapons systems, in our view, it's difficult to envision realistic combat situations where the use of autonomous weapons systems against humans would not pose a significant risk of IHL violations. The challenge of distinction, a risk of algorithmic bias, and the fundamental questions of human dignity make anti-personal systems particularly problematic. Therefore, our delegation supports the proposal to include an additional paragraph on the prohibition of anti-personal autonomous weapons systems. On paragraph six, context appropriate in judgment and control, Sierra Leone strongly supports these provisions at the corner of the regulatory framework. Whatever language is ultimately used, it's essential to retain a formulation that establishes the crucial red line. We believe operators must sufficiently understand and be able to limit the operation and effects of systems such that human deliberative decision making, judgments, and accountability are preserved. On paragraph 7, operationalizing the human control aspect, Sierra Leone welcomes this paragraph and supports the retention of the measures outlined therein, which reflects key components of meaningful human control. We further support the proposal by delegation to replace "should" with "must" for a strong formulation. On 7, we support this provision and support further work for clarity. However, We support the proposal to delete, which includes to avoid conflating distinct forms of restriction. The scale of an operation does not necessarily encompass all of the elements listed after, in particular, the target types. On paragraph 7C, Sierra Leone supports this provision. However, we support the recommendation to remove significantly as it creates unnecessary ambiguity, we believe, regarding the threshold of modification that would trigger application of this provision. In the interest of providing further clarity, we support the proposal for further elaboration of this paragraph. I thank you, Chair.

Let me thank the distinguished representative of Sierra Leone for sharing the views of her delegation and making proposals for changes, amendments. Next on my list is Sri Lanka, to be followed by Bangladesh and the Russian Federation. But let me first pass the floor to the distinguished delegate of Sri Lanka, Madam, you have the floor.

Thank you, Mr. Chair, for giving my delegation the opportunity to share a few additional comments on the Box 3. In general, Sri Lanka would like to mention that the Box 3 is very important in setting rules to prohibit laws which are inconsistent with the international law, particularly with the IHL. While the existing IHL remains a normative framework to govern laws, we believe that it is necessary to explore beyond, particularly a specific legally binding instrument to bridge the gaps and to address the unprecedented challenges posed by lethal autonomous weapon systems. Further, my delegation wishes to propose inclusion of explicit reference to prohibit laws which are inconsistent with IHL and are used without human control and judgment in the box 3. Mr. Chair, Sri Lanka recognizes and supports the centrality of human judgment and control throughout the life cycle of these weapons, and we carefully listen to the different views expressed by the delegations in the room around this concept. My delegation is not fully convinced yet about adding parameters or prefixes to the human judgment and control, which could risk the compliance with the principles of IHL and create accountability gaps. For instance, in the context of extremely high speed attacks, where the context could create a situation of human out of the loop, and in such situations, the term appropriate would risk a narrow interpretation as zero human intervention. Further, due to the speed of the attack, human could do only rubber stamp, and such situations would lead to digital dehumanization. Additionally, depending on the duration, scope, and large volume of data used in an attack, the human judgment and control needs to be an active oversight of specific attacks rather than a one-time authorization. However, understanding the technical and operational aspects of autonomous weapon systems and given the lengthy constructive discussions that we had on this concept over the past sessions, as well as during this session. We remain open to consider the views of the delegations until we reach an agreement as to how we could reflect human judgment and control in the rolling text. Even though we are focusing on legal terms in our work, we believe that the ethical aspects cannot be neglected as laws pose unique and significant ethical implications given the risk of transferring life and death decisions on humans to machines. Therefore, we perceive that reference to ethical consideration is necessary in the text. and we prefer to put it in a stronger formulation in para 7. Further, my delegation would also wish to express our concerns on the reference to incorporate self-destruct, self-deactivation or self-neutralization in para 7, without clarity on the context. Also, we note that self-deactivation option is not an adequate safeguard or solution for fundamental legal, ethical, and security concerns of laws. Thank you.

I thank the distinguished delegate of Sri Lanka for sharing the views of her delegation and making proposals with regard to the rolling text. Next on my list is Bangladesh, to be followed by the Russian Federation. But let me first pass the floor to the distinguished delegate of Bangladesh. You have the floor, sir.

Mr. Chair, thank you for giving me the floor. As my delegation speaks for the first time, we commend your leadership. My delegation would like to comment on the para 7B of box 3, which talks about the limitation of engagements that laws can undertake autonomously. We have heard the different views expressed by the delegations on this aspect. My delegation thinks that this conditioning needs careful consideration by the distinguished delegates. so that it does not lead to unintended consequences in the form of renewed and intense arms race. If past evidence is a guide for us, let me take you to 1970s, when restrictions on delivery vehicles and launchers resulted in intense and unintended arms race for multiple independently targetable reentry vehicle marbles. Since there are no restrictions on number of warheads, states started using single missile payload to host more than one warhead. It made delivery vehicles more lethal and destructive, and therefore destabilizing for strategic stability. One argument for this limitation of engagements may be making mass production of weapons system costly will discourage states from producing them. But historical evidence shows that if a particular weapons system proves to be more effective, states will mass produce them, diverting more scarce resources for this purpose. In the context of laws, That is how states may contemplate avoiding limitation of engagements inbuilt in a particular system if that proves impactful and effective in the battlefield. My delegation therefore draws attention of the delegations to carefully think how best we can reconsider and clarify this part to avoid any unintended consequences. If this limitation of engagement gap is on an individual weapons system, it is highly likely that we will get exposed to that scenario. Thank you, Mr. Chair.

Let me thank the distinguished representative of Bangladesh for also placing our work in a more historical context. Next on my list, and that is for the time being the last request for the floor, let me pass the floor now to the Russian Federation. Sir, you have the floor.

Thank you, Chairman. We'd like to apologize for taking the floor once again, but given the comments we heard from the delegation of the United States, including evaluations regarding the statements that our delegations made, we are now compelled to provide a number of explanations. in order to give the correct impression to all delegations regarding the goals that we have been pursuing in our comments. First of all, in our statement yesterday, as a whole, in any of our statements made by our delegation, so these should not be regarded as criticisms of other delegations or of the positions of other delegations or proposals or suggestions made to other delegations by other delegations. This is not the case. Our comments merely aim to, on the one hand, express some concerns and misgivings which our delegation feels regarding what we hear in the room, and whether we can see that this is leading us to a substantive content that means that the group can fulfill the mandate that was entrusted to it by all states parties and fulfill the tasks given to it. These are the concerns that we share. As this is indeed alarming to us. And We expect the same attitude and understanding in relation to the Russian comments. In relation to our proposals and comments, that there will be no criticism from other delegations either. As if that is the approach that we were to adopt, then we might indeed be able to find a consensus. On our behalf, we are grateful to the statements of the United States for the explanations just provided and for the response that was expressed in answer to our concerns and misgivings. These explanations, if we're frank, do instill some hope in us that this group will be able to fulfill the tasks that it has set to it. We find it hopeful that the delegation of the United States has confirmed that it is committed to previous understandings reached within the framework of this group and the previous reports of this group, as well as the guiding principles, and specifically the understanding that the human element in the interaction between the human and the machine is present, and that human judgment is fundamental... in the context of providing for compliance with the norms of the IHL. And we are of the view... that we do indeed need to use the time remaining to us in order to achieve a proper result Specifically, on the basis of previous outcomes of this group, that is to say, the conclusions and recommendations that were reached, the reports that were adopted, the guiding principles which were approved by States Parties to the Convention, as well as on the basis of the methodology, the structure, and the logic which you had proposed, Mr. Chairman, in the document that we are working on now. If we focus on this document and on the wordings that we have in it, rather than on the full list and the restructurings, and the change to the logic that is thereby achieved, then the chance of reaching a consensus, we believe, is present. And our delegation would like to fully count on that. And once again, we are prepared to reconfirm that we are doing our utmost in order to achieve this goal. You can count on our support. Thank you very much for your understanding.

Let me thank the distinguished representative of the Russian Federation, and especially for his concluding, not concluding remarks, but the final remarks in his intervention. I'm now looking at the list. I have no further requests for the floor. just look around once more, and I think this is it with regard to our discussion on box 3, that in a way was also a much wider discussion also on, on, on our mandate and on how to best fulfill our mandate. So I thank you for all your intervention and your thoughts, also for a number of interesting remarks again this morning with regard to how to address the issues in Box 3. As I stated before, we will now reflect on your remarks and we will work on some language that we can discuss in the second-half of the week. But before we do that, we have, of course, still two other boxes to address. the boxes four and five. And before we do that, I would like to give you a brief update on the informal discussions we had yesterday evening, just to keep everybody fully aware of work that is ongoing. I do not intend to have a further discussion on that, because that is indeed, again, something for later in the week. But I thought that it might be useful, so to bring everybody up to speed, what we have been doing yesterday evening in order to be prepared also for the discussions later in the week when we do a second reading. So what we as team, we listened carefully to you and then started work on 1Alpha. And let me just recall, 1Alpha was inserted after we had changed Power 1 in Box 1. And that was to indeed make sure that the three critical functions, identify, select, and engage, didn't come either to become either overly narrow or overly broad, and that its understanding is sufficiently clear. The insertion of power 1 alpha try to protect, I would say, the interests of two large groupings in this meeting, one grouping concerned about being overly broad, the other one about being overly narrow. Having listened to the reactions on the original subpara 1 alpha, we developed this second one, the new one, the longer one, and there you see two sentences. The first sentence builds on existing language and it replaces the notion of setting the parameters, which was considered to be insufficiently clear and we have replaced it with more specific wording being designs, programs, or otherwise determines a list of potential targets or category of targets. And these words were carefully chosen to encompass all phases in which a human can play a role in the identification process, including in advance of the activation phase. For example, the system can be designed to engage certain types of targets or can be programmed to do so and either scenario should be included. At the same time, we sought to retain the human element in this sentence, which, as we all agree, remains at the core of our work. A list of potential targets or category of targets, in turn, was added to indicate that the target may not just be a single specific target, acknowledging that there may be different kinds of categories of targets, in other words, acknowledging that there is an unknown. in what the system will engage. Lastly, the negative formulation of this sentence is aimed at ensuring that the characterization does not become overly narrow. With the second sentence, the intention is to ensure that systems that clearly should not fall under this characterization are in fact excluded. The word "when" introduces the temporal element to address the engagement moment. While the wording "planning, deciding upon or carrying out" draws on established international humanitarian law terminology from Additional Protocol I. The words "specific target" or "specific targets" were chosen to indicate that in this case the intended target or group of targets is actually specific enough so there is no unknown in what the system will engage. In turn, the second sentence takes a positive formulation, ensuring that the characterization doesn't become overly broad. And I must say we had a very thorough discussion yesterday evening, a very technical, specialist, open discussion, and I would like to thank all those of you who participated in it. And I think it was helpful for perhaps also some of you who were there and did not participate in the discussion, but I think it helped better understanding what is at stake. As always, I would say, even with the Chair's and his team's best efforts, there were still some issues not totally resolved, and there were some questions and remarks. And that led to the following, I would say. Could you put up the next slide? And you see that we worked again on both sentences. We had a discussion on on the part where it went about setting, designing and determining the targets. And we came up with a set of characteristics of potential targets to try and cater for remarks we had with regard to what we initially proposed. And that, because that in a way was developing, that it became ever, ever more longer. And you see already at this moment that 1-alpha is about as long as para-1 itself. And then we move back to a more general description of the designing and determining of the potential targets. That does not exclude the system from the characterization. Still, to cater also for the idea that they might not be concrete targets, but categories we have in certain characteristics of. With regard to the second part of the sentence, we had discussions on whether we were creating loopholes or not, and that led to a change in the very last part of the sentence, where we initially had "deciding upon or carrying out the attack, select the specific targets to be engaged by the system", to make it more specific by saying "carrying out the attack, chooses the specific targets to be engaged by the system". I understand, especially for those of you who were not present yesterday evening, that this is not an easy to swallow morning snack. So for that reason, we already provided it to you now, and we will be able to send it around, I think. Yeah, so we will do so. So we'll give you some time to chew on that during the day. Uh, and we continue our discussions, hopefully, somewhere tomorrow with regard to, uh, also, uh, tomorrow, Friday, on, on box three, but, um, I just wanted to share this with you, um, because this is one of the main issues we really have to address. Uh, I stated at the beginning of, of the week that, um, we have to, to move forward, uh, with regard to getting convergence on, on box one. And I think this is one of the main issues we are facing in Box 1. And we hope this is a positive contribution to that discussion and that we can take this further later this week. So with that distraction, I would now again go back to what we had programmed to do this week and what we had programmed to do today. And that is move forward with the first reading of all the boxes. And we are now starting with box four. Let me give you a brief update with regard to the changes that we have made with regard to box 4, if you compare it with the text we discussed last September. Para 1 and the chapeau have been reworded to address the concern that the previous wording diluted existing legal obligations. as well as the concern on varying international law obligations that exist among our contracting parties. With this slightly revised formulation, I aim to address both points. We would, however, welcome drafting suggestions if your delegation would see that additional changes to the text are necessary to make this clear. Para 2 is not new. It used to be para 4, which I have moved up. as it is directly linked to para 1, with some slight revisions, including the addition of "significant" as a qualifier in order to avoid requiring a legal review for every minor change in the weapons system, as well as matching the language to the characterization for consistency purposes. Para 4 was already present in the previous version of the rolling text, and it illustrates ways for which States should ensure that the capabilities and limitations of laws are understood. It now also includes two sub-paras, alpha and bravo, which are also not completely new. They are slight reformulations of previous paras 5 and 8, from which duplicative or unnecessary redundant parts have been removed. Then about the paras 5 and 6, they retain language on bias mitigation obligations. I inserted the idea of prevent instead of detect and correct to reorient the obligation upstream and make sure measures are applied so biases are not only detected once the weapon systems are developed, as well as making it cumulative, as there should not be a choice to prevent or mitigate. The removal of language on AI is not to imply that this should not be covered, but to avoid creating a loophole whereby mitigation measures are only required for systems that incorporate AI. During our informal consultations, delegations especially stressed the need to find a way to differentiate between elements that are existing legal obligations and those which are not. Moreover, some delegations expressed support for reintegrating several notions of the previous version of the rolling text, in particular the concepts of predictability, reliability, traceability and explainability. And I stand ready to discuss that, and some of you have already referred to it in their interventions earlier. And I would now like to invite delegations to take the floor to provide observations and solu- and solutions. The floor is open. I recognize the distinguished representative of Norway. Let me pass the floor to you, and thank you for breaking the ice.

Thank you, Chair, and thank you for the revised box 4. At this point, I'll just speak on paragraph 1 in the chapeau.

So.

We have made this point a couple of times before, so this should be familiar. We appreciate the emphasis on legal reviews in paragraph 1, yet we find it problematic that legal reviews, which is a treaty obligation for many of us, is couched in non-prescriptive terms through the modal verb "should" in the chapeau. In our opinion, the issue is not resolved by the addition of the words "in addition to their legal obligations" in the chapeau, which, under one interpretation, could seem to reinforce the impression that the paragraphs that follow are in addition to legal obligations and not legal obligations per se. We recognise that box 4 contains elements that are not legal obligations and that many would object to couching those in prescriptive terms, and therefore we do not propose to change the modality in the chapeau itself. Rather, we propose to solve the issue by reordering the paragraph so that the emphasis is on the voluntary exchange of best practice. Our proposal is that the paragraph reads: "engage, as appropriate, in the voluntary exchange of relevant practices between states in order to facilitate legal reviews in accordance with their legal obligations under international law, and to determine in the study, development, acquisition or adoption of laws whether its employment would, in some or all circumstances, be prohibited by international law. That is it for now. Thank you.

I thank the distinguished representative of Norway for making this practical suggestion that would possibly fix the issue. Next on my list is the European Union. Let me pass the floor to the distinguished representative of the European Union, Madame Jaffre.

Thank you, Mr. Chair. With regard to the risk mitigation measures presented in box 4, the European Union supports the inclusion of legal reviews and voluntary exchange of relevant practices between states, conducting testing and evaluation, and ensuring appropriate guidance and training, as well as preventing unintended and automation bias among risk mitigation measures that States should apply prior to potential deployment and as applicable throughout the entire life cycle of LOVs. Thank you.

I thank the distinguished representative of the European Union for stating the Union's positions. Next on my list is the distinguished representative of Japan. You have the floor, ma'am.

Thank you, Mr. Chair, and good morning, colleagues. On section 4, we basically have no problem with the content, but would like to propose to fine-tune some languages. I have three points. The first is in relation to the legal obligation and non-legal obligations. As we explained in the informal consultation, we would like to propose to delete, in addition to their legal obligations, from this chapeau, and instead insert in paragraph 1, "in accordance with their obligations under international law," or alternatively, insert "consistent with their respective obligations under international law." But we remain open to hear other proposals from delegations. Secondly, we support the proposal by Singapore to add "strongly" before "encourage" in paragraph 1. And thirdly, on paragraph 2, we support the proposal by France to modify "significant" to "substantive." We remain flexible on the language, but we are of the view that we should clarify that not all minor technical updates and changes necessitate review. At the same time, we took note of the point made by the US delegation in the informal consultation that we should not be establishing a new standard of significant impact and look forward to further discussion on this point. Thank you, Mr. Chair.

I thank the distinguished representative of Japan for making concrete proposals with regard to the text. And I now pass the floor to the distinguished representative of the Russian Federation.

You have the floor.

Thank you, Chairman. On box 4, we have a number of comments which to a large degree aim to remove duplications, repetitions in this section as compared to other sections. So these are issues already addressed in other boxes. We also aim to consolidate the text in this box. With regard to the chapeau, as the delegation of Japan, which just spoke before us, We also believe that it is necessary to remove the words "in addition to their legal obligations". As in the sub-paragraphs which follow, there are a number of obligations that are already in effect internationally and which affect our countries. And in this connection, the wording which we have in the chapeau does not seem appropriate. That's why we would propose that it should be deleted, since it is not in keeping with the paragraphs that then follow. We believe that we would lose nothing by removing those words. As to paragraph 2, we are of the view that this paragraph is a repetition of paragraph 1 of this box, since the practice of legal reviews presupposes that they are carried out including with regard to any further development or modifications, whatever they may be, to existing weapon systems. So, paragraph 2 is a specific case of paragraph 1, in fact. And in our view, it does not add anything substantively in comparison to what we already have in paragraph 1. Therefore, the need to conduct legal reviews when developing or acquiring or modifying existing weapons systems is an irremovable part of the obligations of Article 36 to the Geneva Conventions, or which is observed voluntarily by a number of states which have not acceded to that additional protocol. And therefore, we propose that paragraph 2 be entirely removed. With respect to paragraph 3, we regard this as a repetition of paragraph 6 of box 3. If you read that paragraph carefully, you can see that there is already mention of the need to ensure human judgment and control over lethal autonomous weapons systems. in order to comply with international law, including international humanitarian law. In box 4, paragraph 3, this is exactly the same concept, that is to say the necessity to provide for such human judgment and control throughout the life cycle of laws. We see no difference between these two paragraphs, and we believe that box 3, paragraph 6 is more developed. It is more comprehensive, and in our view, is genuinely a fundamental understanding of the human judgment that we have in the text in regard to such weapon systems throughout their life cycle. And therefore, we would propose that in box 4, we remove paragraph 3 entirely due to the absence of any added value in it. Paragraph 4B. We propose a small editorial addition. As for a long period of time and including in this document, we have seen the use of the term "chain of command and control". So we would propose that we add the words "and control" after the word "command" in order to be in line with the terminology that we have been using in this document. And in conclusion, on paragraphs 5 and 6, As previously, we believe it is necessary to merge these two paragraphs into one, relating to bias. And in this connection, we believe it could be done fairly simply. We could just add at the end of that paragraph the phrase, and automation bias. And in this way, this paragraph would cover any bias which could arise in relation to the use of these weapon systems and which give rise to concern amongst delegations. So, in this case, we could fully remove paragraph 6, given the addition which we are proposing for paragraph 5. Thank you.

I would now like to pass the floor to the Ambassador of Chile, Leonor, to deliver a group statement. You have the floor.

Thank you. Thank you very much, Chair. I have the honor to deliver this statement on behalf of the African states that are high contracting parties to the Convention on Certain Conventional Weapons. While mindful of our guidance to engage with the current rolling text in a structured manner, I would like to offer general reflections together with comments on the rolling text. And that's only because it took some time for absolute consensus to be arrived at over this paper. So I hope our approach is not entirely disagreeable. The African Group appreciates the continued leadership and commends the inclusive approach taken in developing the rolling text. We recognize that this text reflects constructive contributions from many delegations and represents important progress in our collective work. As we enter the final year of the GGE's mandate, The African Group believes states have advanced their common understandings sufficiently enough to move forward. The rolling text provides a foundation that can serve as a basis for identifying elements of an instrument for future negotiations. While further work is needed, the structure and substance offer a solid starting point. The African Group is ready and committed to completing this mandate's work and supports moving forward toward negotiations on the basis of the rolling text. Mr. Chair, we welcome the characterization in box 1, particularly the clarification in paragraph 1a. This provision ensures that systems where humans set generalized target profiles and operational parameters remain within the scope of regulation, reflecting a realistic understanding of how autonomous weapons systems function. Without this qualifier, the characterization will be interpreted too narrowly, potentially excluding systems that require regulation. We support retaining and further clarifying this provision. We welcome proposals from the ICRC as to how best to clarify, as they have made constructive suggestions on this element of the text. Notwithstanding the notion of context appropriate, in quote, the applicant group strongly supports the principle in box three, paragraph six, that human judgment and control is essential for compliance with international law, including international humanitarian law. This represents the cornerstone of the rolling test and must be preserved. The measures outlined in paragraph seven operationalizes this principle and reflect key components of meaningful human control. Human operators must remain accountable for the operation and effects of weapon systems, which requires users to understand the systems and apply context-specific weapon system decision-making and judgment. The centrality of the human element in the use of force has been in the core part of the discussions of this group and should remain so. We further welcome the provision in box 4, paragraph 3, requiring context-appropriate human judgment and control across the entire life cycle of laws, which reinforces this essential principle. We view the provisions in box 3, paragraph 7, together with the related elements in box 4, paragraph 4, and box 5, paragraph 3, as interconnected starting points that can be developed into legal rules in order to preserve ethical principles, and legal requirements, and ensure meaningful human control over the use of force. Mr. Chair, the African Group emphasizes that within the concept of context, appropriate human judgment and control, there must be adequate predictability, reliability, traceability, and explainability of systems to those responsible for their use. In this regard, the African Group notes the removal of this language from previous versions of the rolling text. These characteristics are essential prerequisites for meaningful human control and for ensuring operators can make informed legal and ethical judgments. We therefore support the proposal by several delegations to reinsert this important language into the rolling text. The chair of the African Group reiterates that autonomous weapons systems raise profound challenges for international humanitarian law, human rights law, human dignity, and international peace and security. With reports of use of increasingly autonomous systems in conflicts, including on the African continent, the urgency of our work cannot be overstated. These systems must remain under the direct, meaningful control and supervision of humans at all times. We have a mandate to formulate by consensus a set of elements of an instrument. We have made substantial progress. Now we must preserve that. what has been achieved, strengthen key provisions, and complete our mandate, so that the path forward to negotiations on a legally binding instrument is clear. The African Group stands ready to engage constructively during this final year to deliver a meaningful outcome that addresses the humanitarian, legal, and ethical concerns posed by autonomous weapons systems. I thank you all, and I thank you for your indulgence. Thank you.

Let me thank the Ambassador of Sierra Leone for speaking on behalf of the African group. Next on my list, Sweden, Australia, Stop Killer Robots, Belgium and Finland. Let me now first pass the floor to the distinguished representative of Sweden. You have the floor.

Thank you, Chair. From Swedish side, we see the same problem as Norway in box four, but would like to propose another solution to the problem. In order not to weaken existing IHL and to have a similar structure as in other boxes, we suggest to remove the chapeau and include states should in all Paris except for the first Paris first sentence. which should be formulated in terms of "must" or "shall". Since this has been discussed before, we could also accept language along the lines earlier proposed by Australia, or the wording of the guiding principle in para E in para 1 in box 4. I thank you.

Let me thank the distinguished representative of Sweden for making concrete suggestions. I think we now have three options to fix this issue. Next on my list is Australia, to be followed by Stop Killer Robots. But let me first pass the floor to Australia. Madam, you have the floor.

Thank you, Chair. Australia is generally supportive of box four and would like to offer some suggestions for further improvement. There remains a diverse range of opinions on the chapeau of box four. The intention of the chapeau seems to be to acknowledge that the paragraphs below may have different statuses for different states depending on the obligations they have under various legal instruments. Australia suggests the chapeau read, "As applicable across the life cycle of laws, states consistent with and in addition to their respective legal obligations should." Australia's proposal attempts to clarify that some states are already legally required to undertake some of the listed measures, while others are not. Our proposal makes language consistent and removes superfluous language. There is no need to specify prior to deployment or entire life cycle, given the measures are to be implemented as applicable across the life cycle. Sometimes these measures will apply and sometimes they won't, like for example, in the decommissioning stage. We have heard other suggestions today to resolve this issue and will give consideration to these. In respect of paragraph 3 of this box, we note that there appears largely duplicative with box 3, paragraph 6. We agree that context appropriate human judgment and control and the variety of measures set out in box 3 paragraph 7 would be applied across the life cycle of laws. We therefore suggest that the phrase "across the life cycle" be incorporated into box 3, paragraph 6. Box 4, paragraph 3, can then be removed as it does not otherwise materially contribute. Box 3 would therefore contain the detailed discussions of context-appropriate human judgment and control. Chair, in respect of paragraph 4, the integration of subparagraphs and represent a positive change. We would like to suggest that due to this integration, it would be beneficial to relocate the second-half of subparagraph to the chapeau. This would make it clear who needs to understand the capabilities and limitations of the laws through the actions set out in subparagraphs and. We also suggest that the appropriate wording be "responsible chain of command and control" and that "human operators" be removed. This would capture all necessary persons and avoid any implication that there would be operators outside a responsible chain of command and control. The proposed chapeau would therefore read, "Ensure that the capabilities and limitations of laws in their expected use and their anticipated effects in different circumstances are understood by those in the responsible chain of command and control of laws through. Paragraph 4 would then simply read, "Ensure appropriate guidance, training and instructions." On paragraph 4, we suggest the addition of "variable" prior to "operational environments." Operational environments may vary, and so testing and evaluation should be conducted in realistic and variable environments. This would give effect to the chapeau of paragraph 4, which seeks that the anticipated effect of laws be understood in different circumstances. Chair, Australia strongly supports the changes you have made to paragraphs 5 and 6. We would suggest that detect still be maintained in both paragraphs as its inclusions ensures that states act to prevent bias, both initially and in an ongoing fashion, implement detection measures to address biases which may nonetheless arise, and act to mitigate any detected biases. I thank you.

I thank the distinguished representative of Australia for sharing the position of her delegation and coming up with concrete proposals. Next on my list is Stop Killer Robots. Let me pass the floor to the distinguished representative of Stop Killer Robots. You have the floor.

Thank you, Chair. Stop Killer Robots believes this text should be negotiated into a legally binding instrument. Therefore, we recommend that the text, in addition to their legal obligations, be deleted from the chapeau of box 4, as the requirements of this section should become the basis for legal rules. Regarding paragraph 1, Stop Killer Robots notes that weapons reviews are a legal obligation for states under Article 36 of Additional Protocol I to the Geneva Conventions, and that an legally binding instrument on autonomous weapons systems or provisions to prevent harm should become legal requirements. We recommend removing the qualifier "significant" that has been added back into the text in paragraph 2 to avoid unnecessarily permissive interpretations. We welcome paragraph 4 and recommend its retention. Ensuring that the effects of systems in their context of use are adequately understood is a key component of meaningful human control. Adequately understanding the effects of a system also depends on predictability, reliability, traceability, and explainability. If text on these concepts is not restored to box 3, we believe they could be usefully placed here instead. We would caution that testing and evaluation based on realistic simulations, as referenced in paragraph 4a, will only be able to give indications of how simpler, deterministic autonomous weapon systems will behave. Systems looking to incorporate more complexity, many of which we believe could not be subject to meaningful human control, could not reliably be tested in this way. Regarding the references to bias in paragraph 5, we believe that algorithmic bias would be a better term to describe the problem being addressed. Steps can be taken to mitigate bias, but it cannot be eliminated. If biased systems are used in life and death decisions, the consequences for groups that are already marginalised and discriminated against are extremely grave. Racism, sexism, ableism and other societal biases will be lethally reproduced and reinforced. One key reason to adopt a specific prohibition on anti-personnel autonomous weapon systems is to remove the possibility of such biased automated killing. Similarly, automation bias is a threat to meaningful human control over weapon systems and upholding international law. The recognition in paragraph 6 of the need to prevent automation bias is therefore welcome. States should engage in further discussion regarding the points at which this risk would feature in the use of autonomous weapon systems and the measures that could be taken to prevent it. Thank you, Chair.

I thank the distinguished representative of Stopka Roberts for sharing the views of her delegation. Next on my list is Belgium, to be followed by Finland. But let me first pass the floor to the distinguished representative of Belgium. So you have the floor.

Thank you, Chair. So, on box 4, we support the Swedish proposals on the use of the words "must" and "should" as was expressed by the Swedish delegate. We would also like to come back on an issue we brought forward in the former sessions, that is dealing with the risk of diversion of laws to the illicit market. We are assured that preventing and halting that terrorist groups, criminals and other unauthorized actors, preventing that they get their hands on laws is a shared objective for all the CCW high contracting parties. And indeed, we have not heard high contracting parties when we brought this up in the former sessions opposing the addition of a reference to diversion. So, the language Belgium has previously proposed in this regard, also already in written format, is the following: States should ensure the adoption of measures to prevent, halt and address the diversion of laws to unauthorized end-users. We would like to see this paragraph taken up in the next version of the rolling text. Thank you, Chair.

I thank the distinguished representative of Belgium for stating position of his delegation and making proposals to add to the text. Next on my list is Finland, to be followed by Partners for Transparency, Turkey and the Republic of Korea. Let me first pass the floor to the distinguished representative of Finland.

Thank you, Mr. Chair. Regarding the chapeau paragraph, the added reference to legal obligations is welcome. However, we would suggest rephrasing it as follows for greater clarity and precision. Prior to potential deployment and as applicable throughout the entire life cycle of laws and in compliance with international law, stage should. Concerning paragraph four, the intention to bring several elements together and streamline the text is positive and contributes to improved readability and coherence. As a general observation, we find that this box places too much emphasis on bias. Testing of systems to ensure safety should be approached in a more multidisciplinary manner, including interdependencies between systems, interfaces, other systems in the operational environment, and cyber resilience. However, we understand that adding new text at this stage would not be advisable and the elements are acceptable in their current form. Thank you, Chair.

I thank the distinguished representative of Finland for commenting on the text and also showing flexibility. Next on my list is Partners for Transparency, to be followed by Turkey. But let me first pass the floor to the distinguished representative of Partners for Transparency.

Thank you, Mr. Chair. Partners for Transparency stresses the importance of transparency in the context of autonomous weapons as it enhances accountability, prevents impunity, ensures no digital bias against certain groups of civilians, and enhances human control over machines. Thus, Partners proposes the following amendments. First, based on existing transparency measures in other conventional weapons agreements, such as the Convention on Clusters Munitions, we encourage adding text to paragraph 4, point B, concerning information exchange and highlighting best practices regarding training, guidance, and proper instructions for those responsible for autonomous weapons systems and their human operators. Thus, the paragraph would read, "Ensuring appropriate guidance, training, and instructions for those responsible for autonomous weapons systems and their human operators, with a commitment to sharing information among States regarding training approaches and the guidance and instructions provided to them." Secondly, Partners welcomes paragraph 5 and 6 in item 4, which acknowledges the need to take measures to prevent algorithmic bias and propose that states replace the phrase "mitigated as much as possible" with the phrase "the complete eradication of digital bias and algorithmic racism in data sets" with a prohibition on deploying any system that does not demonstrate absolute neutrality towards civilian groups based on race, geographical area, and or religion. Thank you.

I thank the distinguished representative of partners for transparency, for commenting on the text and making proposals. Next on my list is Turkey, to be followed by ROK and Ireland. But first, the distinguished representative of Turkey. Madam, you have the.

Thank you, Chair.

We're happy with the overall.

Direction box 4 is going and we only have two small tweaks to clarify the chapeau and paragraph 2. In the chapeau, you would recall I made some remarks also in the informal consultations. We prefer the formulation instead of in addition to their legal obligations, we prefer the following formulation. in accordance with their obligations under international law. The phrase, in addition to their legal obligations, may be interpreted as suggesting that the instrument establishes obligations that go beyond or are distinct from existing international law. It could be read as implying the creation of new legal obligations without clearly defining their scope or basis. That is the information behind our suggestion. The second tweak that we suggested in the informals, and I would be happy to hear some reactions to it if that is palatable, is to insert something in paragraph 2 to explain what we mean by "significant impact". The proposal that we have is to insert, after the words "significant impact", "in terms of operational performance", to specify the scope of what we mean by "significant impact". The new formulation would read as follows: ensure that any further development of or modifications to existing laws that have a significant impact in terms of operational performance on their target identification, selection, and engagement functions, or on their anticipated effects are reviewed to ensure compliance with IHL. Thank you.

I thank the distinguished representative of Turkey for sharing suggestions with regard to the present version of the text. And next on my list is the Republic of Korea. I now pass the floor to the distinguished representative of the Republic of Korea.

Thank you, Chair, and good morning to colleagues. I have four points to make in box 4. First, with respect to the chapeau and paragraph 1, my delegation remains flexible so we can support either the Japanese proposal or the Norwegian formulation which focus more on the voluntary exchange of practices. Second, we concur with Japan that we need some kind of qualifier in front of the term impact, imperative, whether significant or substantial. Third, regarding para 3, we share the views expressed by the Russian Federation that this provision is redundant in light of the earlier text in box 3, paragraph 6 and 8. In addition, when you read this together with the chapeau of this box, paragraph 3 appears more duplicative as the entire life cycle loss is already in the chapeau and again appears in paragraph 3, the life cycle loss. So while we support the principle that relevant measures should apply throughout the entire life cycle, this concept is already reflected in paragraph 8 of box 3. So in the interest of coherence and avoiding the repetition, we suggest deleting paragraph 3 of the box 4. And as previously proposed by our colleagues from Australia, merging paragraph 8 of box 3 into paragraph 4 of box 2. Lastly, as raised during the informal consultations, we propose reverting to the previous version of paragraph 4. The earlier formulation included the phrase, quote, "to enable a human operator to have a reliable expectation of how loss will perform in the anticipated circumstances of its use," end of quote. Its omission weakens the clarity of the purpose why we are having this testing and evaluation. So we therefore recommend reinstating this language. And also we can support the suggestion made by our Australian and Russian colleagues to refer to the chain of command and control in paragraph 4B. Thank you, Chair.

I thank the distinguished representative of the Republic of Korea for stating the positions of his delegation and making proposals for amendments to the text. Next on my list is Ireland, to be followed by France. But let me first pass the floor to the distinguished representative of Ireland.

Thank you, Mr. Chair. On box 4, from our side, As there remains disagreement on which paragraphs should be legal obligations or not, we think it's best to remain agnostic on that at this point. And we think this determination can and should be made in future negotiations, and we hope it does not hold up our work here. So we don't have a clean solution for the tension that was raised by Norway between the chapeau and paragraph 1. We appreciate the suggestions made by Australia, Norway, Sweden, Turkey and others, and we'll study them carefully. In the informals, we had made a suggestion of our own, which I'll repeat here now. But again, we are conscious that due to the outstanding disagreements, not just in paragraph one, but across box four on what should be legal obligations or not, a clean, easy solution here may remain out of our grasp and can be dealt with in future negotiations. Our suggestion we had made was for the chapeau to read, "Prior to potential deployment and as applicable throughout the entire lifecycle of laws, states, in accordance with their legal obligations, must," and then in our view, it would include at least paragraphs 1, 2, 3, and 4, and we will, again, avoid arguing as to why here and now. And for any paragraphs that may not form part of legal obligations, a new chapeau paragraph further down the box can be inserted where legal obligations and in other measures begin. And this could read, in addition to their legal obligations, states should. So this would help delineate and remove any ambiguity once the status of each paragraph is determined at a later stage. paragraph two, we support deleting the word significant before impact as it is subjective, but we can also support the idea of adding substantive, which is an interesting alternative, and we will also consider the proposal made by Turkey just now. In paragraph three, as mentioned by the Republic of Korea, it does seem to repeat a bit of the sentiment in paragraph eight of box three. However, we have heard that some wish to remove that paragraph, and if that's the case, then we do wish to retain it here as a legal obligation. And for consistency throughout the whole rolling text, I think we should decide whether we say across the life cycle or we mention design development and use for consistency. We appreciate the restructuring of paragraph four and the chapeau and its sub paragraphs. In 4A, we suggest to return to inserting the word "rigorous" before "testing." And as I mentioned yesterday, I think here we have an opportunity to reintroduce the concepts of predictability, reliability, traceability, and explainability, which was supported by a number of delegations yesterday. And we have a few options to do this. We've sent them in writing to your team. And I'm happy to go through them briefly now, if that's helpful for all delegations. And so, but just to reiterate that our initial preference is to retain the previous paragraph 6A of box three of the previous version of the rolling text. But some potential alternatives would be, for example, a new paragraph in box four that could read, states must ensure through testing, evaluation, and verification processes across the life cycle of laws, that the effects of laws are adequately predictable, reliable, traceable, and explainable to those responsible for their development and use. And I know I'm probably speaking a little quickly, but I invite delegations to approach if they want the options in writing. Another option is a subparagraph C in box four, paragraph four. which would read, ensuring that system behavior, including the identification, selection, and engagement functions, is adequately predictable and reliable in operation, and that its outputs and decision-making processes are traceable and explainable. A third option is to amend the chapeau of paragraph four, box four, so that it would read, ensure that the capabilities and limitations of laws their expected use and their anticipated effects in different circumstances are understood and that their identification, selection, and engagement functions are adequately predictable, reliable, traceable, and explainable to those responsible for their development and use. Sorry, and comma, in order to enable effective oversight and compliance with IHL, comma, by, so it finishes with the by as it is now. The last option, I think I'll way to propose, which was the idea to split and introduce reliable and predictable in box 4 in the chapeau, as I mentioned, and then introduce traceable in box 5. But I will make that suggestion when we get to box 5. Lastly, paragraphs 5 and 6, we suggest to delete as much as possible, which we see as a necessary qualifier. And we also think we should retain measures to detect, which was in the previous version of the text. And we can support suggestions made on inserting variable and 4A by Australia, as well as a diversion paragraph by Belgium. Thank you, Chair.

I thank the distinguished representative of Ireland for making concrete proposals and please send them also in writing. Next on my list is France, to be followed by Italy, Switzerland and the United States. But let me first pass the floor to the distinguished representative of France. Please, avail of the floor.

Thank you very much, Chair. In the view of our delegation, the measures of this box 4 must be drafted in such a way that the chain of command and the operators that will apply them fully understand them. So they must be clear and pragmatic. And that's already the case in the text that you've submitted. Nevertheless, some minor amendments would still be preferable here. So in paragraph 2, The reviews of lawfulness must focus on those systems that have had significant amendments so that their effect or their framework of use has been changed. It is not relevant, however, to carry out the reviews of any amendment to the system or modification to the system. So, as the Republic of Korea has said before us, we remain flexible on the qualifier which best denotes this, be it significant or substantive. But we certainly do want a qualifier to remain in there. In paragraph 4, we wish to explain that in the view of our delegation, the level of understanding required differs depending on the entities, the context. There must be a reasonable understanding of the systems given the context of their use. In paragraph 4a, this is translated by testing, including appropriate simulations depending on the envisaged use. And in paragraph 4b, this is translated by what must be applied throughout all of the chain of command, depending on the role and function of such a chain. We can therefore support Australia's proposal on these two paragraphs and these two subparagraphs and their chapeau. I'm not going to be coming back on the chapeau of box 4. We've already heard plentiful good solutions suggested by other delegations in order to distinguish between obligations and other measures. Thank you, sir.

I thank the distinguished representative of France for making comments and suggestions and referring to the other suggestions already made with regard to the chapeau. We have a rich choice to make there, I think. Next on my list is Italy, to be followed by Switzerland. Let me first pass the floor to the distinguished representative of Italy. You have the floor.

Thank you, Chair, for giving me the floor and allowing me to comment on this section. Regarding paragraph 2, we support the current formulation, which emphasizes the need to ensure compliance with IHL, to review tools applied during the development or modification of weapons that have a significant impact on the essential functions of LOS. The removal of the previous reference to human judgment and control appropriate to the context, what described in section three, is acceptable to us, considering the application of Article 36 of Additional Protocol I to the Geneva Conventions when assessing new weapons or significant modifications. On paragraph four in the chapeau, We would prefer the addition of the word "adequately" as suggested by others, or similar qualifiers so as not to unduly constrain command actions. In subparagraph A, to clarify the meaning of the adjective "realistic", We suggest replacing it with the phrase, "reflecting actual battlefield conditions to evaluate their performance." This wording is more technical and better reflects operational reality, in our view. About subparagraph C, we agree with the comments provided by the Russian Federation and the Republic of Korea, and we suggest to change the reference to human operator that could create some misinterpretations. We would therefore propose ensuring appropriate guidance, training, and instructions for personnel involved in the responsible chain of command and control of laws. Also, we propose the inclusion of a new subparagraph reflecting point F of the guiding principles, which we consider a security clause for the development and use of LOAs. It would read, when developing or acquiring new weapon systems based on emerging technologies in the area of LOAs, physical security, appropriate and physical safeguards, including cybersecurity against malicious cyber activities, the risk of acquisition by terrorist groups, and the risk of proliferation should all be considered. We presented this proposal already in the previous session, and we did not hear any specific objections. We deem its content to be of utmost importance. Finally, for paragraph 5, we note that the text from the previous version has been unpacked to enhance clarity. It now lists a range of actions aimed at preventing harmful or unintended effects, including the need to account for unintended bias, even in data sets. And I thank you, Chair.

I thank the distinguished representative of Italy for commenting on the text and making proposals. Next on my list is Switzerland, to be followed by the United States and the ICSC. Let me first pass the floor to the distinguished representative of Switzerland.

Thank you, Mr. Chair, for giving us the floor on box 4. This box is important as it sets out practical measures that help ensuring compliance with international law, in particular IHL, and also help implementing the provisions in box 3, regulating permissible systems through meaningful human involvement. Especially if we work towards a legally binding instrument, the chapeau is too weak as it stands now. The term "should" has a consequence that all measures of box 4 are a simple recommendation, especially also context-appropriate human judgment and control. More generally, this puts into question that certain aspects are, as a matter of fact, already a legal obligation, at the very least for a very large part of states. This is, for instance, the case for legal reviews. We therefore propose to delete the reference to "in addition to their legal obligations", as well as "should", the latter of which could be replaced, for instance, with "must". With regard to legal reviews, once the boxes are removed, an option could be to have that as a standalone paragraph without the chapeau. if necessary with previous language of the guiding principle, which stated in accordance with states' obligations under international law, in the study, development, acquisition, or adoption of a new weapon, means, or method of warfare, determination must be made whether its employment would, in some or all circumstances, be prohibited by international law. Regarding paragraph 2, We have a question to what further development exactly refers to. Does it refer to the further development after the initial legal review? If this is the case, this could be explicitly mentioned. And one last point, as Ireland just proposed, It might be a good moment, as we take the text forward, to consider whether we should consistently refer to "across the life cycle" rather than "development and use" or "design, development and use".

I thank the distinguished representative of Switzerland for making proposals and commenting on the text. Next on my list is the US, to be followed by the ICSC, Spain, Portugal, and Austria, but first the US. So you have the floor.

Thank you, Mr. Chairman. Thank you also to other delegations for their good comments and proposals. It's a very rich discussion. Some of the proposals we heard, we probably won't be able to accept, but we appreciate the constructive spirit and look forward to studying these proposals and working with everyone to find what consensus is possible. We have a few overarching comments on box four and we'll provide more detailed written comments. Our primary concern with box four is how it frames the measures specified in it. We can support articulating good practices that help implement specific IHL rules that help mitigate civilian harm or that reduce the risk of unintended engagements. However, we cannot support developing new requirements or norms, especially when we have had these types of systems for many years and no such requirements or norms have been needed. Regarding the chapeau, we support the proposals to delete in addition to their legal obligations. But we've heard the concerns and we are open to working with delegations that are parties to the first additional protocol of 1977 to find a solution that will work for them and for states that are not parties to that instrument. In addition, we view the measures as being applied at specific parts in the life cycle rather than continuously throughout the life cycle. As a result, we propose the following formulation for the chapeau. Prior to potential development and when relevant in the life cycle of laws, states should Regarding paragraphs 1 and 2, it is important to specify that legal reviews are to be conducted of new weapons, an idea that is reflected in Guiding Principle E, which a number of delegations have read out in our discussion today. It is not necessary to conduct a legal review of a loss unless it is new. Similar to our comments on box three, we also believe it is important to clarify that each state has different obligations under IHL. Legal reviews relate to the state's own obligations and not the obligations of other states. Therefore, the first sentence of paragraph one should read, "Conduct legal reviews to determine in the study, development, acquisition, or adoption of a new laws, whether its employment would, in some or all circumstances, be prohibited by the state's obligations under international law. Further, in paragraph two, As we mentioned in the informal discussions, we do not support the creation of a significant impact standard, which is bound to lead to confusion and debate. We appreciate the Russian Federation's recognition that paragraph 2 is related to and included in paragraph 1. However, there may be value in an explicit paragraph that highlights that the modification of systems can trigger the obligation to conduct legal reviews or the practice to conduct legal reviews. Therefore, our proposal is to revise paragraph 2 to read: Ensure that any further development of or modification to existing laws that would make it a new type of laws is also reviewed. Regarding paragraph three, as we have consistently mentioned, we have a fundamental objection to the terminology, context appropriate, human judgment and control, and its placement here. We propose deleting paragraph three. Regarding paragraph four. As mentioned at the outset, we do not support creating abstract new standards of understandability. We therefore propose deleting paragraph 4. Subparagraph is already covered in box 5, and we have made proposals on the CHAPO and subparagraph in the past to address our concerns by connecting these concepts to existing IHL rules that govern attacks, but those proposals have not been accepted. We are open to further discussion with other delegations to address our concerns and to try to find consensus. We are also open to working on language on testing and evaluation in the context of the legal review of weapons. It is a good practice for the results of testing and evaluation to inform legal reviews. Regarding paragraph 5. We believe it needs to be related to the implementation of specific IHL rules. As a result, we would add the phrase at the end connecting the risk of bias that may increase the likelihood of the loss of civilian life, injury to civilians, and damage to civilian objects. We appreciate the Chair's explanation just now for why the reference to AI was deleted. That was one of our questions, so we appreciate the Chair clarifying this just now. We had originally understood this paragraph to address bias in AI capabilities. So removing the reference to AI for us creates uncertainty. Without the reference to AI capabilities, the formulation seems to us to relate to bias in persons rather than bias in machines. We find bias in persons to be better addressed in box five. Here I would also provide an initial reaction to the proposal by the Russian Federation to add automation bias to this paragraph. Again, we would find this edit confusing because we have understood this paragraph to relate to bias in machines, while we understand automation bias to refer to a bias that occurs in persons. So our proposal that we had prepared, you know, before this discussion, before listening to others, you know, was to revise paragraph five to read, "Implementing measures to prevent and mitigate as much as possible unintended bias in AI capabilities, including bias in data sets that may increase the likelihood of the loss of civilian life, injury to civilians, and damage to civilian objects. However, we appreciate the good discussion and the clarification by the Chair just now about why AI was deleted, and we are open to working with other delegations on this language. I think our recommendation is to focus this paragraph on addressing bias in machines while addressing automation bias in box 5. That is also a rationale for deleting paragraph 6 here, and we will outline our proposal on animation bias when we discuss box 5. I thank you, Mr. Chair.

I thank the distinguished representative of the United States for commenting on the text and making proposals and reacting to the proposals of other delegations. Next on my list is the ICSC, to be followed by Spain, Portugal, Austria and Israel. Let's see whether we can make that before the lunch break. But let me first pass the floor now to the distinguished representative of the ICSC. You have the floor.

Thank you, Mr. Chair, and thank you to you and to all the delegations who have spoken on this box this morning. You've saved us from making a much longer intervention because many colleagues have raised many of the points we wish to address on box four. In particular, the question about the chapeau and should versus must and properly reflecting the appropriate legal obligations and so on in a clear way. There have been a number of suggestions made which would meet the concerns, so we're happy to leave it in your hands. What we did want to comment on here is on paragraph six on automation bias. We very much welcome the reference to automation bias here. But we do suggest broadening the scope to cover harmful cognitive biases in human interaction with laws, not just automation bias. There are other kinds of cognitive biases, for example, action bias, which is the tendency to prefer taking action over inaction, even when it's not warranted or justified, which can equally distort human judgment and decision making. So we'd suggest just a minor change to broaden this formulation a little bit, something like implement measures to prevent and mitigate as much as possible harmful cognitive biases in human interaction with laws such as automation bias. We think this would more accurately capture the range of risks and better support compliance with IHL. We've listened to the comments from both the distinguished colleague from the Russian Federation and from the United States about whether to have the two paragraphs five and six or to merge them. At first, what the Russian Federation suggested made sense because it does seem rather repetitive to have those two there. On the other hand, what our colleague from the United States just mentioned, that these are two kind of different categories of bias, the bias inherent in the machine and then the bias relating to human interaction with the machine. So perhaps it's worthwhile keeping those separate. But again, this is something we're open to others' ideas. Thank you very much.

I thank the distinguished representative of the ICSC for commenting on the proposals made. Next on my list is Spain, to be followed by Portugal. But let me pass for now to the distinguished representative of Spain, Madam.

Thank you very much, Chair. My first words must be words of thanks to you and your team for the work and everything you've done so far. The proposals that you've made in this box 4 are acceptable for our delegation. However, we would nonetheless like to make some comments and support proposals made by other delegations. which we understand could help to bring positions closer and improve the text. Now, the introduction that you presented seems to us correct, although there have been different versions presented by delegations, and we will assess all of these proposals. We understand that we must refer to legal obligations and that we should keep "should". On paragraph 1, We don't have problems with this formulation. However, we will consider proposals such as the one submitted by Norway. In paragraph 2, the inclusion of the word "significant" makes it ambiguous and leaves it open to subjective interpretations. As Japan said, it would perhaps be better to have clearer language. we could accept the change of significant to substantive or go along with the proposals made by Turkey. On paragraph 3, my delegation, as we've said at previous statements, believes that a context-appropriate human judgment and control is a good inclusion, but this could be repetitive with the introduction and certain paragraphs in the previous box, as other delegations have pointed out. In paragraph 4 and the two subparagraphs, we believe this is a positive step forward. In subparagraph a, relating to realistic, we believe the proposal made by Italy is sensible. And in subparagraph b, we believe there should be a mention of the change of command and control as an addition, as others have also said. We have a positive view of the proposal made by Australia, which says that this is a proposal or an addition that should be made in the chapeau of four. On paragraphs 5 and 6, now we do take into consideration the comments made by other delegations and the question of strengthening the section on bias and dealing with them independently. However, we believe paragraphs 5 and 6 should be merged. To conclude, we support the proposal made by Belgium on the inclusion of a paragraph including the mention that these systems should not be made available to other actors, including terrorist groups. Thank you, Chair.

I thank the distinguished representative of Spain for commenting on the text, reflecting on proposals made by other delegations. Next on my list is Portugal to be followed by Austria. But let me first pass the floor to the distinguished representative of Portugal. You have the floor.

Thank you, Chair, and thank you also for the text presented. We are generally in agreement with box 4. We would like just to offer some suggestions regarding three aspects of the text. So the first one regarding the chapeau, we also are of the view that a refinement of the language could be useful. We heard many good suggestions in that direction. We could support the majority of them, but we would like to highlight the one presented by Australia, which in our view, it's elegant ensures clarification and a good balance between different positions. Regarding paragraph 2, and specifically on the expression "significant impact", we understand, of course, what this expression is trying to do here in this context. However, the concept is used across different areas of international law in a somewhat ambiguous meaning. And in what concerns IHL, its introduction could really give rise to, in our opinion, to a loophole. So we see different possibilities here, but perhaps the most clean would be just to delete "significant." So, of course, Article 36 of Protocol I only refers to new weapons, but as it was the option here, we also agree that we cannot stay, just keep within a binary concept, whether it's new or it's not. because the most important thing here is really that the weapon, the system, always complies with IHL and particularly with the principles of distinction, proportionality, and precautions. So it really doesn't matter whether the impact is significant or not on all those stages of the life cycle of a weapon, what is important is whether the impact there is impact in terms of the application of international humanitarian law. And when assessing if there is really an impact on targeting, et cetera, that assessment also includes necessarily an assessment on the compliance with IHL. So perhaps here, just deleting significant would be perhaps the easiest and cleanest solution. We would also in this paragraph suggest to delete or engagement, so to, sorry, to change and engagement for or engagement because the impact can be, does not need to be on the three notions here, but just it surprises if it is just on one of them. So, the suggestion is to change "and engagement" for "or engagement". Finally, on paragraph 5, and regarding the expression "unintended bias", we are not sure about this word "unintended" because unintended bias could also violate IH obligations of international humanitarian law. We feel that the distinction should be rather between permissible and unpermissible or lawful and unlawful bias. This is actually in accordance with human rights law, because it's really the unlawful, the unpermissible bias that is based on gender, race, religion, et cetera, that is not permitted. So we would suggest to change the word "unintended" for either "unpermissible" or "unlawful." Thank you.

I thank the distinguished representative of Portugal for commenting on the text making proposals. Next on my list, and I think last before lunch break, will be Austria, and then we still have Israel, Cuba, Brazil, Singapore, and India. But let me first now pass the floor to the distinguished representative of Austria. You have the floor.

Thank you very much. While this box in general is in a very good shape, we also have a few comments and also we want to react to some of the, um, discussion points we heard. When it comes to the chapeau, uh, we, uh, originally when this was issued in December, we had some, some problems understanding the meaning and the purpose of in addition to the legal obligations. I think this could be a problem also since this is a text at some point that will also be read by people outside of this room. So I think we should look into another solution while we understood the original idea and what it was supposed to resolve. The most easy and straightforward solution that we would see would just to have two different chapeaus and one that would address the must obligations and one that addresses the should obligations. We also have no problem of using exactly the same chapeau with different verbs. As Ireland, I mean, we think that the mass obligations could not only encompass one, but also two and four. I will come to three later in the text. And just to start at the beginning, on para one, I think this is highly important that we have Article 36 properly reflected here. I think the addition of new before laws would change it to some extent because Article 36 is meant in a very broad sense. I mean, we are applying it here to those kind of weapon systems, but we would not want to limit it to new laws in every circumstance, and this is an important point. To resolve the issues that some delegations have who are not a party to Additional Protocol I, We could go back to, to the solution that you had in one of the previous texts, texts, which included a phrase like, um, um, in accordance with the legal obligations under international law, I think that's what was one solution we had, um, in the past, um, when it comes to… Two, we were convinced by the excellent comments by Portugal that significant could have a negative impact here as well and that we have to take Article 36 very seriously. The whole idea behind this is also built up on very good work by Australia and other delegations on the impact that new technologies have on Article 36 and how it is applied. And this threshold, when are we talking about a weapon that requires a new legal review, is important, and we need a proper text here. And insofar, we support the expert comments by Portugal on this. We are also looking into the Turkish proposal, which could also be something that could be supported. When it comes to para 3, we Some delegation have linked this to box 2 and 3. For us, this element is highly important and it should be in box 3 at the very end as an issue that relates to either the whole box 3 or to some specific important paras there. If we have it in box 3, which is our default option, and here as well, we would have the problem, we would have the same idea in two different boxes, and maybe this one would be a little bit less ambitious, and it could be a kind of a conflict of interpretation at some point. So we would prefer a very strong development design element in box 3, which would not necessarily require then it having it in a shorter version here in box 4. On box 4, we had some proposals today that could work and that we will definitely look into, which would make the text a little bit more clear and elegant. One big question is, of course, who are we talking about? Who has to implement those measures? And we will look into the Australian proposal. The changes that were made from the last version, which made the whole text more coherent and more clear, also made us lose some elements that we thought are important and that maybe in a revised version could also be reintroduced. I mean, the shortening was done to make some repetitions go away, but we also lost, for example, in A, the reliable expectations of how laws will perform, and in B, we lost the proper understanding of the capabilities and limitations. And we would just ask the Chair's team to look into a way how this can be reintroduced and maybe even put together with some of the proposals by Australia in a meaningful way. On five, we really believe that five and six have to be separated because the issue of bias is a different issue from automation bias, and they also require different measures, and we should not conflate them in any way. On five itself, on the current version, we are looking into-- we also added now including bias in data sets. The point that we already made in the informal consultations is that the datasets are not the only way how bias can enter an AI model. It can also happen through training and programming by humans. They can introduce the bias they carry into the system, and it is then another way of entering, and that's why we would suggest also to add at the very end the formulation and AI models, which would then read, including bias in datasets and AI models.

And we also believe that the word.

Unintended is not really necessary, and this is more or less something that was invented here in this room and not too well known outside of the discussions that we are having here. On the new paras that were proposed, we definitely support the proposal by Belgium to have a new para on diversion. We are also looking favorably towards the proposals that we heard about from Finland and others about having a para that looks into the safety of the systems, maybe building on guiding principle F on cybersecurity. but also maybe including some of the very good proposals that we heard in the past from Ukraine about data poisoning, jamming, and spoofing, which are equally important. And lastly, we would also repeat our proposal to have a new para also that relates to data in general, which would read, "Ensure the integrity, quality, and veracity of data used in the development and training of laws." And that's all.

Thank you.

I thank the distinguished representative of Austria for sharing the positions of his delegation and making proposals. We, yes, as I explained before, we now have still Israel, Cuba, Brazil, Singapore, India, and Canada asking for the floor. So we will address them after lunch. One or two Technical remarks. I would like to kindly inform you of the side event organized by Stop Killer Robots, which will be held today from 1:15 to 2:30 in Room 11, Building A of the Palais des Nations, the same room we were in yesterday evening. this event will provide opportunity for states to reflect on what ground the GGE on laws has covered etc etc and possible further action um let me also repeat that now I already thought it it's so it's in room 11 building a you will find detailed information on the web page of the 2026 session of the GGE on laws I wish you a good lunch break and see you back at 3:00. The meeting stands adjourned.