What happened: A draft U.S. procurement guideline would force AI vendors seeking civilian government contracts to grant an irrevocable licence for the government to use their models for “any lawful” purpose, according to a Financial Times report cited by The Economic Times.

Why it matters: If adopted, the rule would harden “model access” into contract law: firms could be pushed to weaken usage limits they consider safety-critical, and to disclose when models are tuned to meet non‑U.S. compliance regimes — a big deal for both competitiveness and governance.

Wider context: The draft lands amid a public standoff between the Pentagon and Anthropic, after the U.S. military reportedly barred contractors from using Anthropic’s tech over supply‑chain concerns and disagreements about safeguards — signalling a broader shift toward tighter government control of AI procurement.

US draws up strict new AI guidelines amid Anthropic clash: Report — The Economic Times

Singularity Soup Take: Governments want AI on tap without vendor vetoes — but “any lawful use” is a blunt instrument that risks turning safety guardrails into optional extras, exactly when procurement power could be used to demand stronger, auditable controls.

Key Takeaways:

• Contract leverage: The draft guideline would require vendors to grant an irrevocable licence for the U.S. government to use AI systems for any legal purpose, shifting the default from vendor-controlled use policies to state-defined permissibility.
• Neutrality clause: Contractors would be barred from intentionally encoding partisan or ideological judgments in model outputs, a requirement that sounds simple but is notoriously hard to define and test in probabilistic systems.
• Disclosure pressure: Firms would need to reveal whether models were modified to comply with non‑U.S. regulatory frameworks, creating a paper trail that could influence trust, export decisions, and future contract eligibility.