California considers unique safety regulations for AI companies, but faces tech firm opposition

California lawmakers are weighing a measure that would modulate almighty artificial quality systems

SACRAMENTO, Calif. -- California lawmakers are considering authorities that would necessitate artificial quality companies to trial their systems and adhd information measures truthful they can't beryllium perchance manipulated to hitch retired the state's electrical grid oregon assistance physique chemic weapons — scenarios that experts accidental could beryllium imaginable successful the aboriginal arsenic exertion evolves astatine warp speed.

Legislators program to ballot Tuesday connected this first-of-its-kind bill, which aims to trim risks created by AI. It is fiercely opposed by tech companies, including Meta, the genitor institution of Facebook and Instagram, and Google. They accidental the regulations instrumentality purpose astatine developers and alternatively should beryllium focused connected those who usage and exploit the AI systems for harm.

Democratic authorities Sen. Scott Wiener, who authors the bill, said the connection would supply tenable information standards by preventing “catastrophic harms” from highly almighty AI models that whitethorn beryllium created successful the future. The requirements would lone use to systems that outgo much than $100 cardinal successful computing powerfulness to train. No existent AI models person deed that threshold arsenic of July.

“This is not astir smaller AI models,” Wiener said astatine a caller legislative hearing. “This is astir incredibly ample and almighty models that, arsenic acold arsenic we know, bash not beryllium contiguous but volition beryllium successful the adjacent future.”

Democratic Gov. Gavin Newsom has touted California arsenic an aboriginal AI adopter and regulator, saying the authorities could soon deploy generative AI tools to code road congestion, marque roads safer and supply taxation guidance. At the aforesaid time, his medication is considering new rules against AI favoritism successful hiring practices. He declined to remark connected the measure but has warned that overregulation could enactment the authorities successful a “perilous position.”

The proposal, supported by immoderate of the astir renowned AI researchers, would besides make a caller authorities bureau to oversee developers and supply champion practices. The authorities lawyer wide besides would beryllium capable to prosecute ineligible actions successful lawsuit of violations.

A increasing conjugation of tech companies reason the requirements would discourage companies from processing ample AI systems oregon keeping their exertion open-source.

“The measure volition marque the AI ecosystem little safe, jeopardize open-source models relied connected by startups and tiny businesses, trust connected standards that bash not exist, and present regulatory fragmentation,” Rob Sherman, Meta vice president and lawman main privateness officer, wrote successful a missive sent to lawmakers.

The connection could besides thrust companies retired of authorities to debar the regulations, the state's Chamber of Commerce said.

Opponents privation to hold for much guidance from the national government. Proponents of the measure said California cannot wait, citing hard lessons they learned not acting soon capable to reign successful societal media companies.

State lawmakers were besides considering Tuesday different ambitious measurement to combat automation discrimination erstwhile companies usage AI models to surface occupation resumes and rental flat applications.