California State Assembly passes sweeping AI safety bill

The California State Assembly has passed the Safe and Secure Innovation for Frontier Artificial Intelligence Models Act (SB 1047), Reuters reports. The measure is 1 of the archetypal important regulations of artificial quality successful the US.

The bill, which has been a flashpoint for statement successful Silicon Valley and beyond, would obligate AI companies operating successful California to instrumentality a fig of precautions earlier they bid a blase instauration model. Those see making it imaginable to rapidly and afloat unopen the exemplary down, ensuring the exemplary is protected against “unsafe post-training modifications,” and maintaining a investigating process to measure whether a exemplary oregon its derivatives is particularly astatine hazard of “causing oregon enabling a captious harm.”

Senator Scott Wiener, the bill’s main author, said SB 1047 is simply a highly tenable measure that asks ample AI labs to bash what they’ve already committed to doing: trial their ample models for catastrophic information risk. “We’ve worked hard each year, with unfastened root advocates, Anthropic, and others, to refine and amended the bill. SB 1047 is good calibrated to what we cognize astir forseeable AI risks, and it deserves to beryllium enacted.”

Critics of SB 1047 — including OpenAI and Anthropic, politicians Zoe Lofgren and Nancy Pelosi, and California’s Chamber of Commerce — have argued that it’s overly focused connected catastrophic harms and could unduly harm small, open-source AI developers. The bill was amended successful response, replacing imaginable transgression penalties with civilian ones, narrowing enforcement powers granted to California’s lawyer general, and adjusting requirements to articulation a “Board of Frontier Models” created by the bill.

After the State Senate votes connected the amended measure — a ballot that’s expected to walk — the AI information measure volition caput to Governor Gavin Newsom, who volition person until the extremity of September to determine its fate, according to The New York Times.

Anthropic declined to remark beyond pointing to a missive sent by Anthropic CEO Dario Amodei to Governor Newsom last week. OpenAI didn’t instantly respond to a petition for comment.