2026 Group of Governmental Experts on Lethal Autonomous Weapon Systems
First Session (Geneva, 2–6 March 2026)
Asia-Pacific Institute for Law and Security
Utrecht University
Thank you, Chair!
I intervene once again on behalf of the Asia-Pacific Institute for Law and Security (APILS) delegation and my own, Utrecht University.
We see Box IV as containing a range of important compliance measures, and we appreciate the efforts, Chair, of you and your team, to better organise and streamline the language in this box.
Our most fundamental concern with this box, also shared by others who have intervened today, is that it does not sufficiently distinguish between legal obligations and other measures. This blurring occurs as a result of the conduct of legal reviews being placed under a chapeau that used the modal verb “should”. Many delegations have underlined that the obligation to undertake legal reviews of new weapons constitutes a legal obligation at least for those states that are party to Additional Protocol I. We have noted this as well, and expressed concern that the inclusion of text whereby legal reviews “should” be conducted risks diluting this obligation.
In this version of the rolling text, this issue has become exacerbated by the introduction of the phrase “in addition to their legal obligations” in the chapeau, and by the removal of the phrase “in accordance with their obligations under international law” from paragraph 1.
Chair,
We recognise that different states are in a different legal position when it comes to the obligation to conduct legal reviews. The language of the rolling text must therefore be sensitive to differing legal obligations but not simply adopt the lowest common denominator.
We note that the group’s earlier consensus documents have managed to resolve this problem. For example, paragraph 23 of the 2023 Report noted that:
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.
This formulation accommodates both those states that are under an obligation to conduct legal reviews and those states that are not.
Furthermore, paragraph 2 of Box IV concerns modification of existing LAWS by requiring a renewed review to ensure compliance with applicable law. While Article 36 of Additional Protocol I does not explicitly use the term “modification”, it is generally understood in State practice that modifications that may affect a weapon’s characteristics or effects trigger the need for a new or updated legal review. Accordingly, paragraph 2 also reflects an aspect of an existing obligation, for those states that have this obligation, rather than a supplementary commitment.
For these reasons, prefacing paragraphs 1 and 2 with the clause with “in addition to their existing legal obligations” is inconsistent with existing law and must be avoided.
Additionally, as others have mentioned, we recommend removing the term “significant” as it may narrow the scope of the paragraph. We believe this term lacks clarity as to the threshold, much in line with the discussion on para 7C in Box III yesterday.
Chair,
We recommend that the substance of paragraphs 1 and 2 be moved before the current chapeau, and reordering the language for a better flow. We further recommend that these paragraphs draw on the consensus language of the 2023 report, while substituting “a new LAWS” for the more generic terms “a new weapon, means or method of warfare” that was used in 2023.
Thus, we propose the following three paragraphs, all to be placed at the beginning of Box IV, before the current chapeau. I will read them out slowly and provide the text to the secretariat:
In accordance with States’ obligations under international law, in the study, development, acquisition, or adoption of a new LAWS, a determination must be made whether its employment would, in some or all circumstances, be prohibited by international law.
This also applies to any further development of, or modifications to, existing LAWS that have a significant impact on their target identification, selection, and engagement functions or on their anticipated effects.
In this context, the voluntary exchange of relevant best practices between States is encouraged, bearing in mind national security considerations or commercial restrictions on proprietary information.
Chair,
In subparagraph 4(A), which was previously paragraph 5, the removal of the word “rigorous” creates a materially weaker compliance measure. The word “rigorous” signals a certain qualitative threshold and implies that testing should be methodologically sound, systematic, sufficiently extensive, stress-tested under varied conditions, and so on. Without that qualifier, the clause merely requires that some testing and evaluation occur, potentially transforming the testing requirement into a procedural formality.
In the same subparagraph, we suggest retaining the clause that such testing is “to enable a human user to have a reliable expectation of how LAWS will perform in the anticipated circumstances of its use” as it explicitly connects testing to an outcome. It is rigorous testing that makes a “reliable expectation” possible, not just any testing. Moreover, this proposal is consistent with the suggestion of multiple delegations to reflect the notions of predictability and reliability in the rolling text.
In paragraphs 5 and 6, a reference to the detection of bias has been lost. Removing “detect” eliminates an explicit compliance measure to monitor and audit systems for bias. We should look at “detect” as an activity aiming at actively identifying bias through testing and auditing, “correct” as an activity actively remedying identified flaws, and “mitigate” as activity directed at reducing residual risk. The revised version significantly weakens the commitment to this three-part dynamic lifecycle oversight (monitor–correct–mitigate).
I thank you, Chair!
