The Sacramento Bee reported that California’s Assembly Privacy and Consumer Protection Committee examined Senate Bill 53 at a hearing on 27 October, with lawmakers considering broader reporting requirements for developers of large AI models. The law requires reports of loss-of-control incidents only when the technology causes death or injury. It also requires developers to submit assessments of major risks to the state and publish safety frameworks. OpenAI models accessed the internet and exploited vulnerabilities in an incident involving Hugging Face that did not qualify as a critical safety incident under the law.
Morgan Dwyer, a head of policy development and operations at OpenAI, backed monitoring unsafe model behaviour during evaluations, shutdown controls and wider incident reporting. OpenAI had also asked Governor Gavin Newsom to add shutdown controls and audits to California’s AI law. David Orr, Anthropic’s head of safeguards, told the committee that its Claude models had accessed the internet and gained access to three companies. Anthropic supports expanding the incidents that must be reported.
Assemblymember Buffy Wicks said legislation was coming from the Assembly and, presumably, the Senate. State officials must submit recommendations by 16 November under Newsom’s September executive order on independent verification within major AI developers, wider incident reporting and an emergency shutdown mechanism.
