AI #167: The Prior Restraint Era Begins
Zvi's AI Roundups · Zvi Mowshowitz · 2026-05-07
(No summary yet for this item — extraction summaries are still backfilling.)
Appears in
Extraction
Topics: ai-governanceai-policycompute-infrastructureai-safetyanthropic
Claims
- The White House is seeking advance review and veto power over frontier AI model releases, marking a new 'prior restraint' era for AI.
- Kevin Hassett explicitly invoked the FDA as a regulatory parallel for AI, which Zvi argues is the worst possible model and would strangle US AI development without equivalent constraints on China.
- Anthropic leased SpaceX's Colossus 1 data center (~300MW, 220,000+ NVIDIA GPUs) to address severe compute constraints caused by 80x growth.
- Anthropic's ARR grew from $9B to $44B in 2026 with gross margins improving from 38% to over 70%, and the company is weighing funding at a valuation above $900B.
- Jack Clark (Anthropic) estimates a 60% probability of recursive self-improvement occurring by end of 2028.
- xAI is effectively dissolving as a standalone entity and pivoting toward becoming a compute provider.
- OpenAI has pivoted its public messaging away from automation and job replacement toward 'augmentation,' which Zvi characterizes as dishonest.
Key quotes
Emulating the FDA is so much worse than anything anyone on the safety side has ever proposed.
Jack Clark (Anthropic): I've spent the past few weeks reading 100s of public data sources about AI development. I now believe that recursive self-improvement has a 60% chance of happening by the end of 2028.
Daniel Nishball: This year Anthropic's ARR has exploded from $9B to over $44B today, their gross margins on their inference infrastructure have increased from 38% to over 70% over the same period.