Project Meridian and the race to govern military AI
As the US builds an AI-first military with its own autonomous warfare command, who is setting the limits on military AI, and where?
Civil AI laws such as the EU AI Act stop short of military uses, so limits on military AI have come from three other places: lab usage policies, US executive orders and procurement terms, and slow UN talks. In January 2026 the Pentagon demanded ‘any lawful use’ terms from AI suppliers. In February it labelled Anthropic a supply-chain risk, and on 30 September it launched Project Meridian. UN talks on binding rules resume in November.
KeyStoryRule or lawAlso on another Storyline
- UNIDIR warns AI is crowding out human judgment in war and opens a forum on military AI
UNIDIR says AI already crowds out human judgment in war and opens a UN forum to write voluntary guidelines for the companies building it.
- Hegseth launches Project Meridian with Musk, Luckey and Gingrich to plan future war
Musk, Luckey and Gingrich are asked to define the capabilities of future war in 120 days.
Also on The Pentagon, the labs and the limits on use, Who supplies the AI-first military?
- UN expert group agrees draft final report on autonomous weapons; treaty question deferred
Geneva talks end in a non-binding consensus report. The treaty question is left to November.
- Presidential memo orders rapid rewrite of Pentagon autonomous-weapons rules
The White House orders a fast rewrite of the main US autonomous-weapons directive.
- Rule, In forceUSEO 14409 (covered frontier models)
Washington’s frontier-model security order relies on voluntary pre-release access instead of licensing.
Also on The Pentagon, the labs and the limits on use, Frontier AI and cyber attacks, What frontier AI developers must disclose
- Pentagon labels Anthropic a supply-chain risk after it refused to drop two limits
Anthropic refuses to drop its limits on surveillance and autonomous weapons.
Also on The Pentagon, the labs and the limits on use, Who supplies the AI-first military?, Procurement as AI policy
- Hegseth orders an ‘AI-first’ military and ‘any lawful use’ terms for AI contracts
The Pentagon demands ‘any lawful use’ terms and new models within 30 days of release.
Also on The Pentagon, the labs and the limits on use, Who supplies the AI-first military?
- UN General Assembly backs autonomous-weapons resolution 164-6; US votes no
164 states back work toward an autonomous-weapons instrument, and the US is one of six no votes.
- UN General Assembly votes 166-3 for talks on lethal autonomous weapons in 2025
The UN General Assembly’s December 2024 resolution on lethal autonomous weapons came a year before resolution 80/57.
- Rule, Partly in forceEUEU AI Act
The world’s broadest AI law excludes systems used exclusively for military, defence or national security (Art. 2(3)).
Also on What frontier AI developers must disclose, Who decides if a medical AI is safe?
- UN General Assembly adopts resolution 78/241 on lethal autonomous weapons
The General Assembly’s 2023 resolution put autonomous weapons on its agenda.
Related developments
Connected developments not on this line
- DevelopmentPentagon says it has stopped using Anthropic’s products5 Oct 2026, Procurement, United States
- DevelopmentAppeals court upholds Pentagon’s supply-chain-risk label on Anthropic25 Sep 2026, Enforcement or court case, United States
- DevelopmentOpenAI agrees to deploy its models on Department of War classified networks27 Feb 2026, Deal, United States
- DevelopmentDepartment of War launches GenAI.mil with Google’s Gemini for Government9 Dec 2025, Launch, United States
- DevelopmentPentagon AI office awards Anthropic, Google and xAI up to $200m each, after OpenAI’s June award14 Jul 2025, Procurement, United States
- DevelopmentCCW states give the autonomous weapons expert group a three-year mandate to 202617 Nov 2023, Policy, International