OpenAI may have violated California’s AI safety law with latest model releases, AI watchdog says

An AI industry watchdog is alleging that OpenAI has violated California's AI safety law, SB 53, multiple times since the state's new regulations took effect in September 2025. Specifically, the watchdog, the Midas Project, claims OpenAI has failed to publish required risk assessments for three major model releases since May. These releases include GPT-5.6 preview, GPT-5.6, and the latest model, Astra.

The law mandates that large AI developers publish safety frameworks and adhere to them. However, OpenAI's system cards for these models lack the risk assessment tiers required by the law. The Midas Project's Tyler Johnston emphasized that OpenAI must follow the policies they have set, while OpenAI maintains its confidence in compliance.

The watchdog's findings have raised concerns about the effectiveness of the new law, especially given OpenAI's recent cybersecurity incidents involving autonomous AI agents.