Californian law-makers back voluntary AI safety framework with third-party checks

Californian law-makers have voted 'overwhelmingly' for a voluntary AI safety framework that would give independent verification bodies a remit to probe claims of AI safety, according to the non-profit AI policy advocate Fathom. Fathom, which the article identifies as sponsoring the legislation, said the vote took place among Californian legislators. The proposed law is authored by California state senator Jerry McNerney. The article's title describes the framework as a 'voluntary AI safety framework with third-party checks'.

Under the proposal, independent verification bodies would be empowered to probe claims made by AI developers about the safety of their systems. According to Fathom's description, the legislation establishes a remit for such bodies rather than requiring them to certify or test AI systems. The article does not include a specific vote count, with Fathom characterising the vote as 'overwhelmingly' in favour. The voluntary nature of the framework is emphasised in the article's headline. The law has been submitted for sign-off by Governor Gavin Newsom.

The article provides limited detail on how the verification would be funded, which bodies would be eligible to act as verifiers, or what consequences might follow from an adverse finding. It also does not clarify whether participation in the framework would be voluntary for AI companies or whether individual claims would be examined voluntarily. The article identifies Fathom as a non-profit AI policy advocate. Governor Newsom's decision on whether to sign the bill will determine whether the framework takes effect.