Box IV - Statement by Stop Killer Robots to the GGE on lethal autonomous weapons systems, 2-6 March 2026

The following statement was delivered on Wednesday, 4 March 2026, by Lauren Barnard to delegates participating in the meeting of the Group of Governmental Experts on Lethal Autonomous Weapons Systems (LAWS), which took place on 2-6 March 2026.

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 1 to the Geneva Conventions, and that in a legally binding instrument on autonomous weapons systems all 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 usefully be 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 weapons 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 weapons systems is to remove the possibility of such biased automated killing.

Similarly, automation bias is a threat to meaningful human control over weapons 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 weapons systems, and the measures that could be taken to prevent it.

Thank you, Chair.