Les engagements du Canada en matière de gouvernance de l'IA et les systèmes de ciblage militaire par l'IA des alliés fonctionnent sous des hypothèses divergentes

Canada has stated that fully autonomous weapons systems would be unacceptable and that the Canadian Armed Forces is committed to maintaining appropriate human involvement in the use of military capabilities that can exert lethal force. At the 78th UN General Assembly in December 2023, Canada voted in favour of Resolution 78/241 on autonomous weapons, and at the 79th session in December 2024, voted in favour of Resolution 79/239 on lethal autonomous weapons systems.

Allied militaries with which Canada must maintain interoperability are deploying AI targeting systems that operate under different assumptions about the speed and nature of human oversight.

In February 2026, the United States used the Maven Smart System — integrating Anthropic's Claude via a Palantir contract and running on AWS — in Operation Epic Fury against Iran. The system helped generate approximately 1,000 strike targets in the first 24 hours. The U.S. conducted approximately 900 strikes in the first 12 hours. Reporting described how AI-driven systems compressed decision cycles from weeks into minutes.

In Gaza, the Israel Defense Forces used AI targeting systems Lavender and The Gospel. An investigation by +972 Magazine documented Lavender's approximately 10% error rate and analyst review time of approximately 20 seconds per target. Human Rights Watch published separate analysis of the legal and methodological concerns about the use of machine learning for targeting. The AI Incident Database catalogued this as Incident 672.

Concurrently, DND/CAF's own AI strategy, drafted in 2022 and not approved until March 2024, acknowledged that neither DND nor the CAF is "positioned to adopt and take advantage of AI" and described AI initiatives as "fragmented, with each command and environment addressing AI independently." In October 2025, Canada's National Security and Intelligence Review Agency (NSIRA) initiated a formal review of AI use in national security and intelligence activities, indicating that AI deployment in the security apparatus has outpaced existing oversight frameworks.

The interoperability hazard is structural: in a coalition operation, CAF personnel may need to act on intelligence, targeting data, or operational plans generated by allied AI systems that operate at speeds and error tolerances that are incompatible with Canada's stated policy on human oversight. No framework currently exists to manage this gap.