OpenAI Offers A New Policy Blueprint
Zvi's AI Roundups · Zvi Mowshowitz · 2026-06-05
Zvi Mowshowitz analyzes OpenAI's proposed federal AI governance blueprint, which calls for CAISI-centered mandatory model evaluations and federal preemption of state AI safety laws, finding the document surprisingly reasonable while warning sharply about enforcement risks and preemption scope.
Appears in
Extraction
Topics: ai-governanceai-policyrecursive-self-improvementfederal-regulationai-safety
Claims
- OpenAI's blueprint calls for CAISI to conduct mandatory evaluations of frontier models without authority to block deployments.
- OpenAI publicly acknowledges early signs of recursive self-improvement in current AI systems and calls for RSI to be treated as an urgent governance priority.
- The blueprint proposes building a national framework on California's SB 53, New York's RAISE Act, and Illinois's SB 315, with federal preemption of state frontier safety laws in exchange.
- Mowshowitz identifies federal preemption as the most dangerous element of the proposal, warning enforcement may never materialize once state leverage is surrendered.
- OpenAI proposes concentrating oversight authority in CAISI rather than the NSA to prevent a de facto licensing regime.
Key quotes
We also see early signs of recursive self-improvement (RSI) in today's systems: where AI development is itself accelerated by AI. We expect this to increase competitive pressures among developers and nations, and create governance challenges that existing institutions are not equipped to address.
This is a highly reasonable document, well exceeding expectations.
If we accept preemption of all 'frontier risks' this risks meaning that states can enact counterproductive laws about AI but not the ones that matter, and there is the risk Congress never again acts.