Lines of Thought

The Pentagon, the labs and the limits on use

Who decides how the US military may use frontier AI: the government that buys it or the companies that build it?

In 18 months the US defence relationship with frontier labs went from $200m pilot awards to a fight over usage limits that ended in court. This line follows the contracts, the 'any lawful use' demand, the Anthropic exclusion and OpenAI's classified deal, through to the ruling and Project Meridian.

  1. ProcurementDefenceUS
    Pentagon AI office awards Anthropic, Google, OpenAI and xAI up to $200m each

    The starting point: four labs get Pentagon awards of up to $200m each.

    Also on Who supplies the AI-first military?

  2. ProcurementGovernmentUS
    OpenAI, Anthropic and Google offer frontier AI to US agencies for about $1 via GSA

    The same labs offer frontier AI to all federal agencies for about $1.

    Also on Procurement as AI policy

  3. PolicyDefenceUS
    Hegseth orders an 'AI-first' military and 'any lawful use' terms for AI contracts

    The department demands 'any lawful use' terms in every AI contract within 180 days.

    Also on Project Meridian and the race to govern military AI, Who supplies the AI-first military?

  4. Enforcement or rulingDefenceUS
    Pentagon labels Anthropic a supply-chain risk after it refuses to drop two limits

    Anthropic keeps two limits, on mass domestic surveillance and fully autonomous weapons, and is cut off.

    Also on Project Meridian and the race to govern military AI, Who supplies the AI-first military?, Procurement as AI policy

  5. DealDefenceUS
    OpenAI agrees to deploy its models on Department of War classified networks

    A day later OpenAI signs a classified-network deal with its own stated limits written into the contract.

    Also on Who supplies the AI-first military?

  6. Rule · In forceUS
    EO 14409 (covered frontier models)

    Washington's frontier-model security order relies on voluntary pre-release access rather than licensing.

    Also on Project Meridian and the race to govern military AI, When models learned to hack, Transparency, not licensing

  7. Enforcement or rulingDefenceUS
    Appeals court upholds Pentagon's supply-chain-risk label on Anthropic

    An appeals court rules that a vendor's usage restrictions can count as a supply-chain risk.

    Also on Who supplies the AI-first military?, Procurement as AI policy

  8. ProgrammeDefenceUS
    Hegseth launches Project Meridian with Musk, Luckey and Gingrich to plan future war

    Five days later the department hands long-range technology planning to a study led by industry figures.

    Also on Project Meridian and the race to govern military AI, Who supplies the AI-first military?

Where this line could go next

Connected developments not on this line