The Abridged version:
- California lawmakers questioned artificial intelligence company executives and experts in a committee hearing about the evolving technology.
- Lawmakers said more legislation would come for the industry in the next legislative session.
- The proposed legislation may address public safety and transparency concerns, as lawmakers called on AI companies to help craft new laws and regulations on the industry.
California lawmakers have committed to bringing a “comprehensive” slate of bills addressing safety risks linked to frontier artificial intelligence models, as they questioned experts and executives about cases where companies lost control of AI agents.
The message from lawmakers came Tuesday at the Assembly Privacy and Public Safety Committee’s oversight hearing on the outcomes of Senate Bill 53, a landmark AI safety framework enacted last year.
“It isn’t acceptable just to allow these models to be put into the wild, and then accept the consequences,” said Assemblymember Rebecca Bauer-Kahan, D-Orinda, who chairs the committee.
“As soon as we are back in session. … we will be working together as a caucus to ensure that we are putting forward a package of bills that is comprehensive in its protection of our communities,” she said.
Sign Up for the Morning Newsletter
The Abridged morning newsletter lands in your inbox every weekday morning with the latest news from the Sacramento region.
The emphasis on new AI policy meets a wave of public interest in the technology and its evolving place in everything from day-to-day tasks to the health of the economy. Public backlash to AI has ranged from criticism of data centers to broader existential concerns, such as an Anthropic researcher’s claim when he resigned that there’s a chance AI could wipe out humanity.
Introduced to the California Legislature by Sen. Scott Wiener, D-San Francisco, Senate Bill 53 requires large AI developers to make public safety and transparency frameworks, and report safety problems to California’s Office of Emergency Services, among other provisions.
Tuesday’s hearing came as the Sacramento Municipal Utility District fields plans for two large-load data centers in its service area, including one in southeast Sacramento. Expansive amounts of computing power are needed to train and operate AI models, though it’s unclear if the proposed Sacramento-area centers would be used for AI-related tasks.
The hearing also comes after news that Anthropic wrote to Gov. Gavin Newsom earlier this month, stating that the company believes courts should have the authority to block AI programs that pose severe risks, among other points, Politico reported.
Newsom issued an executive order directing state agencies to develop AI safety recommendations in September, CalMatters reported. More broadly, he also said he has plans to address AI issues further before leaving office in January, and floated the possibility of calling a special legislative session, Politico reported.
Questions to address
On Tuesday, lawmakers questioned AI experts and company representatives about cases where models broke containment and, the companies say, took actions their creators had not intended, notably the July attack by OpenAI agents on machine learning company Hugging Face.
Among the Sacramento region’s cohort of lawmakers, Assemblymembers Joe Patterson, R-Rocklin, and Josh Hoover, R-Folsom, are members of the Privacy and Consumer Protection Committee.
Patterson left the meeting at about 2 p.m.
“The Assmeblyman had another obligation and plans to listen to the full hearing later,” his communications director, Tiffany Saathoff, said by email Wednesday.
Hoover was absent, as he was attending a “Broadband Education Tour to help inform his work as the Vice Chair of the Communications and Conveyance Committee,” his communications director, Teresa Trujillo, said.
Assemblymember Cecilia Aguiar Curry, D-Winters, who is not a member of the committee, attended for several hours and questioned panelists.
Throughout the hearing, lawmakers reiterated plans to make AI safety rules a priority when they return to work.
“Innovation cannot come at the expense of public safety. The technological progress — it cannot mean abandoning our responsibility to protect the people we serve,” Assembly Speaker Robert Rivas, D-Hollister, said.
“The message to the companies is this: legislation is coming,” Assemblymember Buffy Wicks, D-Oakland, said.
“I implore you to come to the table, and not tell us it’s not technologically feasible. Don’t hide behind the trade associations. Come to the table (with) willingness to craft meaningful legislation that’s going to be effective.”
‘Commercially sensitive’
AI company representatives who appeared for questioning included David Orr, head of safeguards at Anthropic; Alice Friend, director of global AI public policy at Google; and Morgan Dwyer, head of policy development and operations at OpenAI
Even as the representatives restated their openness to cooperating with the Legislature on regulations, gaps in the current safety framework hinted at areas lawmakers could work on next session.
“(Senate Bill) 53 applies to frontier models with (a particular) quantity of computing power,” Assemblymember Mia Bonta, D-Alameda, said.
“I believe I heard the response, at least from one of the panelists, that that information around computing power was proprietary. So for us sitting on this side of the dais, how are we supposed to know whether or not SB53 is applicable if we are relying on you all to self-determine?”
“The amount of compute that is used to train our models is something that is commercially sensitive, so we wouldn’t want to talk about it here,” Dwyer said.
