State Sen. Josh Becker wants major artificial intelligence developers to disclose the computing resources they devote to safety research and meet state standards for those allocations.
Becker, a Democrat from Menlo Park who represents San Mateo, announced plans Oct. 20 to introduce the Frontier AI Safety Act. The proposal would require quarterly reports on the share of research computing resources used for safety and authorize state officials to set minimum allocation standards.
The California Governor’s Office of Emergency Services would oversee reporting procedures, compliance timelines and verification requirements, according to Becker’s announcement. Violations could trigger civil enforcement by the state attorney general.
“If you’re investing in making AI more powerful, you must also invest in making it safer,” Becker said in the announcement.
The proposed requirements would cover computing resources used to research and develop new AI systems. They would not restrict computing used to provide products and services to customers.
Covered developers would have to obtain independent verification of their reported allocations. Becker’s office said verification would involve organizations under the framework created by SB 813, authored by Sen. Jerry McNerney, D-Pleasanton.
Gov. Gavin Newsom signed SB 813 on Sept. 9. It takes effect Jan. 1, 2027, according to McNerney’s office.
Becker’s office cited Anthropic’s disclosure that about 6% of its AI research and development computing resources went toward safety during one seven-day period.
In its original analysis, Anthropic described a July 13-20 snapshot and cautioned that computing use is an imperfect measure of safety investment because safety research can require less computing than training powerful models.
Becker’s office said other major developers had not publicly disclosed comparable figures.
The proposal would build on California’s existing AI oversight laws. SB 53, authored by Sen. Scott Wiener, established transparency and incident-reporting requirements for major frontier AI developers and whistleblower protections, according to the governor’s office. SB 813 establishes a voluntary framework for independent AI assessments.
Becker expects to pitch the measure when lawmakers return to Sacramento in December for the new legislative session, Politico reported. The publication said OpenAI did not comment on the proposal for its reporting.
