California lawmaker aims to put up guardrails for AI development

SACRAMENTO, Calif. (CN) — A California state senator on Thursday introduced a bill he said would create safety standards when developing artificial intelligence, predicting that it would cement the state’s future as the world’s AI capital.

Senate Bill 1047, called the Safe and Secure Innovation for Frontier Artificial Intelligence Systems Act, would establish what state Senator Scott Wiener calls “clear, predictable, common-sense” standards for the developers of the biggest and most powerful AI systems. The bill doesn’t affect startups that use less powerful AI.

Additionally, the bill would create CalCompute — a cloud-computing space that would be publicly available. It would let startups, researchers and others develop high-level AI.

Wiener in a statement said AI has the potential to help advance knowledge and understanding in fields ranging from medicine to wildlife forecasting. It also gives the state the chance to use the lessons it’s learned about the negative effects the growing technology can have without proper understanding.

“SB 1047 does just that, by developing responsible, appropriate guardrails around development of the biggest, most high-impact AI systems to ensure they are used to improve Californians’ lives without compromising safety or security,” Wiener, a San Francisco Democrat, said.

The state is a center for AI, and Wiener said California must invest in its development across government, business, and civil sectors, ensuring it keeps its place as the leader in AI development.

However, the state also must maintain vigilance about the possible dangers it poses.

People at the forefront of AI have issued warnings about the failure to create guardrails for the technology. AI could endanger public safety and national security without proper development, posing risk to infrastructure, the possible use of biological weapons and cyberattacks.

Wiener’s bill seeks to stop those threats before they happen.

Among other requirements, the law would require a developer of powerful AI to have the ability to completely shut down the artificial intelligence, until it could be deemed safe. In other circumstances, the developer would need to obtain an annual certification of compliance.

Thomas Strohmer, director of UC Davis’ Center for Data Science and Artificial Intelligence Research, said it’s important for the government to take regulatory steps in AI development. He pointed to environmental regulations, noting the government doesn’t leave it to the private sector to regulate that aspect of the business.

“We can’t leave it to tech companies to do the right thing,” Strohmer said.

For Strohmer, data privacy is key. He said there’s a conflict between what tech companies want from people and what the people want for themselves. Laws must ensure people are protected, and that tech giants don’t crush legislation before it becomes law.

The right laws must be tailored properly. AI’s uses include many areas, and Strohmer said the same rules for health care also shouldn’t apply to creating better advertising.

Ensuring innovation can happen is essential to Strohmer. He said Wiener’s bill accounts for this, as its focus is on large developers making powerful AI. Innovation often comes from smaller startups, which aren’t targeted in the bill.

“I think they did take the right approach,” he added. “Finding the right balance is really, really important.”

Wiener introduced the bill the same day the U.S. Department of Commerce’s National Institute of Standards and Technology launched the U.S. AI Safety Institute Consortium, the country’s first entity focused on creating AI measurements and evaluations. Senate Bill 1047 is intended to bring the consortium’s work to California and put its best practices into law.

It also falls on the heels of action taken both by California Governor Gavin Newsom and President Joe Biden.

Newsom in September issued an executive order seeking to shape the future of AI through ethics, transparency and trustworthiness. The governor also wants to keep the state at the top of the AI field.

Biden followed with an executive order the following month calling for the creation of safety standards and privacy protections. It also promotes innovation and competition.

Wiener’s bill is one of several pieces of AI-related legislation that California lawmakers have introduced this term.

California already has a law requiring the creation of an inventory of high-risk automated systems that could be used by the state. Assembly Bill 1824 — introduced by Assemblymember Avelino Valencia, an Anaheim Democrat — would make a disclosure requirement for any content made by AI.

Senate Bill 942 — written by state Senator Josh Becker, a Menlo Park Democrat — would lead to the creation of tool that could let people easily determine if a picture, audio, video or text was made by AI.

State Senator Angelique Ashby, a Democrat from Sacramento, wrote Senate Bill 970. It would require anyone who sells or has access to AI that’s intended to create images, videos and voice recordings to include a consumer warning.

And state Senator Aisha Wahab — a Bay Area Democrat — introduced Senate Bill 933, which would include an image made by AI in the definition of a computer-generated image for the purposes of child pornography.