California Governor Gavin Newsom released a roadmap for implementing the state’s new AI laws that calls for independent oversight of frontier AI. The recommendations include requiring large frontier AI developers to let licensed evaluators assess catastrophic risks, with evaluators reporting to the state rather than the companies they review.
The roadmap also says the state should prioritize assessments of automated AI research and development, including recursive self-improvement. Fathom, an organization that advocates for independent verification, says the report reflects many of its recommendations. The roadmap outlines a proposed framework; it does not establish when the requirements will take effect.
Why it matters
If adopted, the plan would require frontier AI developers to undergo independent risk assessments, but the roadmap does not say when those obligations would begin.