Didem Mente/Getty Images
- OpenAI opposes California's AI bill, arguing that it could drive companies out of the state.
- The bill requires strict safety protocols for AI models, including a "kill switch."
- Despite industry opposition, the bill will be voted on in the California Assembly at the end of the month.
OpenAI is joining Silicon Valley's tech titans in the fight against California's landmark artificial intelligence regulation bill, Bloomberg reported on Thursday.
SB 1047, introduced by California State Sen. Scott Wiener in February, seeks to establish "common sense safety standards" for AI systems that cost over $100 million to develop, Business Insider reported on Monday.
The bill mandates that companies implement protocols to prevent their AI models from causing "critical harms," such as being used in cyberattacks or leading to the development of weapons of mass destruction.
The bill also specified the provision of a "full shutdown," which functions as a kill switch for AI systems.
Jason Kwon, OpenAI's chief strategy officer, warned that the bill could stifle progress and drive companies out of California, in a letter addressed to Wiener on Wednesday.
Kwon also wrote that regulation of AI concerning national security is "best managed at the federal level" rather than through a "patchwork of state laws."
Silicon Valley tech heavyweights like Meta and Anthropic have been lobbying against the bill, too.
Meta warned that the bill might discourage the open-source movement by exposing developers to significant legal liabilities, wrote Rob Sherman, vice president of policy and deputy chief privacy officer at Meta, in a letter in June. Sherman wrote that regulation could hamper the broader tech ecosystem because smaller businesses rely on these freely available models to innovate.
Anthropic also resisted the bill's stringent preemptive regulations, advocating instead for a more balanced approach that wouldn't stymie progress, BI reported on Monday.
OpenAI previously lobbied against similar legislation by the European Union. The company sought to ease the regulatory requirements on general-purpose AI systems like GPT-3, Time reported last year.
The EU had since altered its final draft of the AI Act to exclude language that would classify general-purpose AI as high risk, instead focusing on "foundation models" with more limited requirements, according to Time.
Despite the industry opposition, Sen. Wiener argued that it is a "highly reasonable bill that asks large AI labs to do what they've already committed to doing," the senator wrote in response to OpenAI's letter on Wednesday.
The bill had passed a vote in the state Senate and is set for a final vote in the California Assembly at the end of the month.
OpenAI and Sen. Weiner didn't respond to a request for comment sent outside standard business hours.