OpenAI Offers A New Policy Blueprint

Abstraction: OpenAI's federal frontier-AI safety framework blueprint

Key points:

• OpenAI's "Democratic Governance of Frontier AI" blueprint calls RSI "potentially the most consequential frontier safety issue of the coming decade" and urges CAISI treat it as an urgent priority.
• Three-part strategy: a national framework built on state laws (SB 53, NY RAISE Act, IL SB 315), strengthening CAISI as primary safety institution (not NSA), and a whole-of-government resilience plan.
• CAISI would run mandatory evaluations but explicitly "conduct evaluations and recommend mitigations — not approve or block deployments"; in exchange Congress preempts state frontier-safety laws.
• Zvi is impressed (exceeds expectations) but lists five concerns: fines like $1M are far too weak for meaningful accountability, federal enforcement may simply not happen or be selective, watered-down compromises risk becoming the ceiling, preemption scope must be carefully tailored, and it must not be treated as "problem solved."
• Prefers concentrating responsibility in civilian CAISI and avoiding power concentration in NSA; Nathan Calvin wary of accepting preemption / giving up reverse federalism.

Connections: Zvi Mowshowitz · Openai · Caisi · Nsa · Federal Preemption · AI Governance · AI Policy

Source: https://thezvi.substack.com/p/openai-offers-a-new-policy-blueprint