Military AI Policy by Contract: The Limits of Procurement as Governance

11 hours ago

The U.S. is adopting a flexible but inadequate AI governance model through regulation by contract, bypassing democratic accountability and public deliberation.
The Pentagon's 'any lawful use' policy requires AI vendors to remove restrictions beyond legal requirements, prioritizing speed over governance.
Anthropic was designated as a supply chain risk after refusing to drop its red lines, leading to legal challenges against the government's decision.
OpenAI's agreement with the Pentagon, framed around 'any lawful use,' shifts interpretive authority to the government, raising concerns about enforceability.
Contractual terms for military AI are being publicly rewritten due to public backlash, highlighting the lack of formal governance review.
The 'consistent with applicable laws' clause in OpenAI's agreement leaves enforcement dependent on the government's interpretation, not the vendor.
The 'intentionally' qualifier in OpenAI's prohibition on domestic surveillance may limit its effectiveness by allowing room for interpretation.
Federal contracts, especially Other Transaction (OT) agreements, lack robust enforcement mechanisms, making it difficult to prevent misuse of AI technologies.
The Pentagon's reliance on AI like Claude has reached a point where termination of contracts may not be feasible, undermining vendor control.
The administration's shift toward commercial-first procurement reforms risks removing baseline procedural defaults, placing more weight on bilateral negotiations.

Hasty Briefsbeta