California's ambitious attempt to regulate large-scale artificial intelligence (AI) systems is moving forward in the state Legislature, but with significant revisions and fierce debate within the tech industry.
Senate Bill 1047, introduced by Sen. Scott Wiener, D-San Francisco, passed the Assembly Appropriations Committee with significant amendments last week. The bill aims to establish safety standards for robust AI systems amid growing concerns about the potential risks of unchecked AI development. Industry experts worry that provisions in the bill could stifle innovation.
“Companies driving AI innovation need a regulatory environment that balances safety with their ability to grow and develop cutting-edge solutions,” Daniel Christman, co-founder of AI company Cranium, told PYMNTS. He warned that excessive regulation “could stifle the natural advantages of generative AI and stifle the innovative potential these models offer.”
Changes in Enforcement
The original draft of SB 1047 included possible perjury penalties for non-compliance, a provision that drew sharp criticism from AI developers. The revised bill now relies solely on civil penalties.
“We can promote both innovation and safety; the two are not mutually exclusive,” Wiener said in a news release defending the bill's progress.
Dev Nag, CEO of software company QueryPal and former CTO of Wavefront, sees the bill's changes as a step in the right direction. “Eliminating perjury penalties will help reduce disincentives for AI in California, but the penalties in their original form didn't work,” Nag told PYMNTS.
Another amendment is the abolition of the Frontier Model Division (FMD), the proposed new regulatory body to oversee the implementation of the Bill. Instead, existing government-run agencies will absorb some of the FMD's planned functions.
Nag expressed concern about the change, saying, “While eliminating the proposed new regulatory agency may streamline some bureaucratic processes, analyzing AI models is not a skill that state attorneys general's offices have ever had to demonstrate or employ.”
The bill's legal framework has also been changed: developers will now be required to follow a “reasonable care” standard instead of the more stringent “reasonable assurance” requirement, a change that brings the bill closer to established legal precedent.
Taking into account the concerns of startups and small AI companies, the revised bill sets a $10 million threshold, meaning only fine-tuned models with costs above that amount will be covered by the bill, effectively exempting many small businesses in the AI field.
Enforcement mechanisms have also been narrowed: state attorneys general's offices can only seek civil penalties if damage has been caused or public safety is under imminent threat, limiting the scope of preemptive measures.
The bill's focus has drawn scrutiny from industry experts, with Nag pointing to a possible mismatch in priorities, saying “the bill itself seems more interested in speculative risks in the distant future than risks that exist today, like misinformation and deep fakes.”
This concern reflects a broader debate within the AI community: A recent survey found that while 70% of AI researchers believe safety should be a priority in AI research, 73% expressed “substantial” or “extreme” concern that AI could fall into the hands of dangerous groups.
Despite these concerns, the bill has garnered support from some prominent members of the AI community, including Geoffrey Hinton, who is well known as one of the “godfathers of AI,” who said SB 1047 “takes a very sensible approach to balancing these concerns.”
California's tech leadership is at stake
The global race to develop and regulate AI has brought California's longstanding tech leadership under scrutiny, with industry experts warning that overly strict regulation could jeopardize the state's competitiveness.
“Maintaining a competitive balance with other states and countries is essential for California to remain a center of AI development, especially as other jurisdictions relax their regulatory environments,” Nag emphasized. He noted California's historical leadership in a range of technology fields, including hardware, software, biotechnology and aerospace, and attributed this success to an innovation-friendly environment.
The tech industry is watching closely as the bill goes to Congress on Tuesday (August 20) for passage by August 31. With federal legislation on regulating AI stalled, California's efforts could set a precedent for other states and shape the future of AI development nationwide.
“GenAI is a lot like a hammer; it's a very versatile tool,” Christman said. “In effect, the state is trying to regulate the hammer manufacturers, not the people who are using them to commit nefarious activities.”
To stay up to date on all things PYMNTS AI, subscribe to our daily AI newsletter.
Read more: AI, AI Regulation, Artificial Intelligence, California AI Bill, Cranium, Daniel Christman, Dev Nag, Frontier Model Division, GenAI, Generative AI, Government Run Entities, News, PYMNTS News, QueryPal, Reasonable Assurance, Reasonable Accommodation, Scott Wiener, Senate Bill 1047
Source link
