California alleges three OpenAI safety violations

The Midas Project alleges OpenAI violated California’s AI safety law at least three times this year by omitting required risk tiers from GPT-5.6 preview, GPT-5.6, and GPT-6 Astra, Fortune reports. OpenAI’s Frontier Governance Framework defines four categories—cyber offense, CBRN, harmful manipulation, and loss of control—while California’s civil penalty can reach $1 million per violation, according to the report. Until OpenAI answers the allegation, builders should request model-level tier records rather than accept a high-level framework, a procurement test that extends the archive’s Astra access-gate analysis.