[Update] Asser Institute at UN Informal Exchanges on AI in the Military Domain

The discussion reflected a growing recognition that AI decision-support systems (DSS) - software platforms that use artificial intelligence, machine learning, and data analytic - raise distinct and urgent questions that warrant dedicated attention. Marta Bo described this as a “welcome broadening of the conversation beyond autonomous weapons systems”.  

Asser Institute researchers have engaged with International Humanitarian Law (IHL) and the accountability issues raised by military AI and AI DSS in targeting. As Taylor Kate Woodcock shows in her recent PhD thesis, sustained reliance on AI can subtly reshape how legal assessments are made during targeting. 

AI shapes both situational awareness and factual assessments, and so influences the discretionary judgments commanders must make regarding IHL principles like distinction, precautions, and proportionality. Among the risks is that AI may narrow the friction points needed for target verification and contextual legal judgement, contributing to a systematisation of harm. 

Marta Bo addressed the question of war crimes responsibility in the use of DSS, drawing on earlier analysis of Project Maven, the US Pentagon's leading AI programme, launched in 2017 to help military analysts filter through vast amounts of drone footage from war zones.  

Bo identified the accountability risks the programme raises – in particular around proof of intent and knowledge, automation bias, and mistake of fact. These concerns have since been underscored by recent reporting on the alleged use of the Maven platform in attacks in Iran, including in the devastating attack on the Minab school. 

Broadening the conversation on AI

Drawing on the UN Informal Exchanges, Marta Bo and Taylor Kate Woodcock identify several areas that should inform and broaden the conversation on the use of DSS in the military domain:

• A deepening of the shared understanding of the specific technical features of DSS and their effects on human agency; 
• Identifying concrete, context-dependent measures to preserve human agency and judgement through dialogue and exchanges with developers and operators; 
• Examining DSS beyond the targeting context, including upstream uses (for example for threat analysis) and effects on jus ad bellum and the threshold for the use of force and uses in broader security contexts; 
• Promoting the consideration of international law throughout the AI lifecycle, including during development, procurement, and review; 
• Exchanging best practices as states develop new legal review procedures and legal assessments for DSS; 
• Clarifying interpretation of key IHL obligations, particularly what additional precautions are required in the use of DSS; 
• Addressing what increased reliance on AI DSS means for individual criminal responsibility for war crimes, and in particular for assessing intent and knowledge; and 
• Addressing pressing questions of corporate responsibility and liability, given the growing involvement of industry in targeting decisions.

Read more

The Asser Institute has longstanding expertise in the international law dimensions of military AI, including AWS and DSS, and conducts research that supports the development and specification of international law obligations applicable to military AI.