By Shirin Ghaffary, Bloomberg
State Senator Scott Wiener, a San Francisco Democrat, has launched a invoice that may require firms growing AI fashions above a sure computing efficiency threshold to publicly launch security and safety protocols that assess the potential catastrophic dangers to humanity from the expertise. Below the regulation, AI firms additionally would wish to report any “critical safety incidents,” comparable to theft of delicate technical particulars, to the state legal professional basic. Firms that could be affected by the proposed laws embrace OpenAI, Alphabet Inc.’s Google and Anthropic.
The primary-in-the-nation transparency necessities are a part of amendments launched Tuesday to SB 53, which Wiener launched earlier this 12 months. The laws contains protections for whistleblowers inside AI firms and the creation of a public cloud to supply low-cost entry to computing energy for startups and tutorial researchers.
“As AI continues its remarkable advancement, it’s critical that lawmakers work with our top AI minds to craft policies that support AI’s huge potential benefits while guarding against material risks,” Wiener mentioned in remarks ready for the discharge of the amendments. “SB 53 strikes the right balance between boosting innovation and establishing guardrails to support trust, fairness and accountability in the most remarkable new technology in years.”
The proposed necessities come after the US Senate lately blocked an effort to bar particular person states from enacting AI rules within the shortly advancing discipline. Wiener’s effort to push regulation regardless of his earlier failed try reveals how AI regulation continues to be a preferred subject amongst legislators who say they need to shield the general public from the expertise’s doubtlessly devastating harms.
Whereas SB 53 has transparency necessities and potential penalties if firms don’t disclose vital security incidents, it doesn’t embrace SB 1047’s provision on legal responsibility.
Already, firms like Meta Platforms Inc., Google, OpenAI and Anthropic, frequently launch security tips for his or her fashions; SB 53 goals to codify and standardize these procedures.
The amended invoice is anticipated to go earlier than a California Meeting committee later this month. Wiener referred to as it a “work in progress” and mentioned he might be speaking with stakeholders within the coming weeks to refine the proposal.
Extra tales like this can be found on bloomberg.com
©2025 Bloomberg L.P.
Initially Printed: July 9, 2025 at 6:25 AM PDT
