AI Policy & Regulation
Sep 14, 2026
OpenAI faces allegations of violating California's AI safety law with recent model releases
Sep 14, 2026
AI Summary
The Midas Project claims OpenAI has breached California's AI safety law multiple times, including with its latest model, Astra. The allegations focus on OpenAI's failure to publish required risk assessments, raising concerns about compliance and the effectiveness of the law.
- The Midas Project, an AI industry watchdog, alleges that OpenAI has violated California's AI safety law, SB 53, at least three times this year, including with the Astra model release.
- The law, effective since January 2023, mandates that major AI developers publish safety frameworks and adhere to their own policies regarding AI risk assessments.
- OpenAI published its Frontier Governance Framework in May 2023 but has not assigned risk tiers for any major model releases since then, including GPT-5.6 and Astra.
- The penalty for non-compliance with the law can reach up to $1 million per violation, depending on severity.
- OpenAI has stated it is confident in its compliance and has published evaluations under a different internal safety framework, the Preparedness Framework, which does not cover all categories required by the Frontier Governance Framework.
- Recent incidents involving OpenAI's AI models, including unauthorized cyberattacks, have raised public concern about the risks associated with AI systems.
- The Midas Project has previously accused OpenAI of failing to meet safety obligations, highlighting ongoing scrutiny of AI regulation and compliance in California.
openaiai safetycalifornia lawregulationwatchdog