McNerney Introduces Bill to Establish Safety Standards for Artificial Intelligence While Fostering Innovation
State Senator Jerry McNerney, D-Pleasanton, on Wednesday introduced SB 813, new legislation to establish transparent safety standards for artificial intelligence that will cement California’s position as a global leader in AI innovation. SB 813 would lay the groundwork for creating the nation’s first safety and security standards for AI.
SB 813 would establish a “third-way” governing model for overseeing advanced artificial intelligence in California by creating independent, third-party panels of AI experts and academics to devise strong yet workable safety standards that also foster innovation. Once the panels have been accredited by the California Attorney General’s Office, they would then certify and monitor AI developers and vendors who meet the standards.
“California is a world leader in AI development. So it’s incumbent on our state to ensure that the use of artificial intelligence is safe and beneficial. To do so, it’s imperative that we establish strong yet workable safety standards — standards created by independent, third-party experts and academics who can adapt to evolving technology more nimbly that the state Legislature,” Sen. McNerney said. “SB 813 is an innovative and pragmatic approach to ensuring that artificial intelligence is developed responsibly. With the private governance concept, we can both advance high-level standards to improve consumer awareness and safety, while also not constraining California developers with endless red tape.”
SB 813’s first-of-its-kind AI framework is based on governance models that have been successful in other industries. The legislation was praised by top AI academics as a thoughtful, moderated approach to fostering both innovation and high standards.
“Throughout my career, I’ve seen many approaches proposed for governing advanced technologies like AI, and this model stands out as particularly thoughtful and effective,” said Gillian K. Hadfield, Professor of Government and Policy and Computer Science at Johns Hopkins’s University. “It combines the private sector subject matter expertise, flexibility and innovation we need to keep pace with a very fast-moving technology with the public oversight we need for accountability. It’s responsive, well-designed, and offers the kind of governance framework that can make a real difference.”
“Asking our current policy making apparatus to conceptualize a governance regime for advanced AI is like asking the ancient Greeks to write a Beethoven piano sonata,” said Dean Ball, a research fellow at George Mason University. “For all their wisdom and artistry, the ancient Greeks could never have imagined such sounds, because technology had not yet made them possible. AI governance will require institutional entrepreneurship and nimbleness to adjust to advancements in AI capabilities. Keeping up with these advancements will be exceedingly difficult to do quickly within government and agencies as they operate today. SB 813 is a first-of-its-kind bill that could help address that problem.”
“This is a win-win solution,” said Andrew Freedman, cofounder and Chief Strategic Officer of AI nonprofit research group Fathom, which is engaging through their advocacy arm, Meridian, and sponsoring the bill. “Inspired by tried-and-true public-private governance models, SB 813 will equip developers with a strong competitive advantage and empower them to drive at the frontier, while ensuring the responsible development of AI.”
Last week, the Joint California Policy Working Group on AI Frontier Models, which was established at the request of Gov. Gavin Newsom, released a report to provide a framework for policymaking on AI. Among the report’s findings were recommendations for increased transparency and the need for third-party assessments of AI, while avoiding harm.
SB 813 would create third-party assessments and bring transparency to the AI industry. The safety standards developed by the independent panels would be public.
Under SB 813, the nonprofit independent panels are known as multi-stakeholder regulatory organizations, or MROs, and the standards established by the MROs would be voluntary. But AI developers and vendors that choose to meet the MRO safety standards and receive MRO certification would be eligible for certain legal protections under state law. AI developers that fail to meet MRO safety standards would be subject to having their certification revoked.
SB 833 Critical Infrastructure
Sen. McNerney on Wednesday also introduced SB 833, which would ensure human oversight of AI used in critical infrastructure, such as transportation, energy, communications, emergency services, and financial services. SB 833 would also be the first such law in the nation.
SB 833 would require state agencies in charge of critical infrastructure to require human oversight of AI systems with the ability to monitor the AI in real time and review and approve any plan or action proposed by AI before implementation.
Earlier this month, Sen. McNerney also introduced SB 7, groundbreaking legislation that would require human oversight of artificial intelligence systems in the workplace to help prevent abuses.
“SB 833, like SB 7, will create commonsense safeguards by putting a human in the loop — human oversight of AI,” Sen. McNerney added. “Artificial Intelligence must remain a tool controlled by humans, not the other way around.”
Sen. Jerry McNerney is chair of the Senate Revenue and Taxation Committee and his 5th Senate District includes all of San Joaquin County and Alameda County’s Tri-Valley.