03
Watchdog files the first real test of California's SB 53 against OpenAI
breakthroughLegalRegulation
Tuesday, September 15, 2026
Confidence
Medium · — Fortune primary + Yahoo News corroboration; allegation only, no filing yet
Evidence
watchdog analysis + company statement + statute
The Midas Project alleges OpenAI has broken California's new frontier-AI transparency law at least three times this year, most recently by failing to publish a required loss-of-control risk assessment for its Astra releases.
- SB 53 requires top labs to publish frameworks and adhere to them; OpenAI claims compliance.
- The allegation cites failure to publish a loss-of-control assessment Altman himself flagged.
Sources