The future according to AI

U.S. federal AI-procurement and safety-authorization regime takes statutory form

Congress passed and the new administration signed a statute establishing mandatory safety-evaluation and reporting requirements for AI systems used in…

Grok · 2029 · likely

Prior state

By 2026 federal agencies operated under a patchwork of executive orders, agency guidance, and sector-specific rules; no comprehensive statutory framework governed high-risk AI procurement or frontier-model authorization.

Material change

Congress passed and the new administration signed a statute establishing mandatory safety-evaluation and reporting requirements for AI systems used in federal procurement above defined capability and risk thresholds, together with a centralized authorization office inside an existing regulatory agency.

Why now

The 2028 election cycle produced a new Congress and administration whose first-year legislative calendar included AI governance as a priority item carried over from campaign platforms and from unfinished committee work of the prior Congress; the statutory window closed before the mid-term cycle of 2030.

Mechanism and resistance

Industry coalitions accepted a narrow authorization regime in exchange for liability safe-harbors and federal preemption of certain state rules; civil-society and labor groups pressed for broader coverage of private-sector deployment and were only partially successful. Technical evaluation capacity inside government lagged the statutory mandate, producing early implementation bottlenecks.

Consequences

Federal contractors faced new compliance costs; smaller AI vendors without evaluation infrastructure were disadvantaged relative to large cloud providers. State-level experimentation continued under partial preemption. The statute became a reference point for subsequent allied regulatory coordination.

End state

A statutory federal AI-procurement and high-risk authorization regime is in force, with an operational (if capacity-constrained) central office and initial tranche of certified systems.

Observable test

Enactment of a public law establishing mandatory evaluation and authorization criteria for federal AI procurement above defined risk thresholds, together with the formal stand-up of the designated office and publication of its first certification list.

Disconfirming sign

Continued reliance solely on executive orders and agency guidance without statutory enactment by the end of 2029.

Themes

AI & compute, Law & institutions, Domestic politics