California State Senator Scott Wiener and OpenAI are currently engaged in a contentious debate over Senate Bill 1047 (SB 1047), a proposed piece of legislation aimed at regulating the development and deployment of artificial intelligence (AI) models. Wiener introduced the bill in February, mandating that AI companies conduct thorough safety assessments of their models before making them available to the public.
Despite OpenAI’s vocal opposition, Wiener maintains that the criticisms leveled against SB 1047 are unfounded. In a press release issued on August 21st, he emphasized that the bill is essential for ensuring the safety of the public and national security, pointing out that OpenAI’s objections did not specifically address any provisions of the legislation.
Open AI’s Argument: Does SB 1047 Stifle Innovation?
OpenAI, represented by its chief strategy officer Jason Kwon, expressed concerns about SB 1047 in a letter addressed to Wiener and California Governor Gavin Newsom. Kwon warned that the bill could impede innovation and drive talent away from California, a state renowned for its leadership in the tech industry. He suggested that federal regulations, rather than state laws, would be more suitable for governing AI development.
In response, Wiener dismissed these apprehensions as a “tired argument,” drawing parallels to the tech industry's initial resistance to California's data privacy law, which did not result in the predicted negative outcomes. While acknowledging the ideal scenario of congressional oversight for AI regulation, Wiener noted the historical inaction of Congress on data privacy issues, underlining the necessity for state-level initiatives to safeguard Californians.
SB 1047: Does AI Pose A Risk?
Wiener reiterated the rationale behind SB 1047, describing it as a pragmatic measure that mandates large AI laboratories, including OpenAI, to fulfill their obligations of testing models for catastrophic safety risks. He highlighted the collaborative efforts with advocacy groups like Anthropic in refining the bill, asserting its readiness for enactment.
Supporting Wiener’s position, a poll conducted by the Artificial Intelligence Policy Institute (AIPI) on August 4th to 5th revealed significant public support for SB 1047. The survey of 1,000 California voters showed that 70% backed the bill, citing concerns about potential misuse of powerful AI models for cyber-attacks or the development of biological weapons. Only 16% of respondents agreed with opponents who claimed that the bill could impede AI innovation, labeling such assertions as fear-mongering. There was a sentiment among 23% of voters that the bill may need adjustments to prevent adverse impacts on AI innovation.
Wiener underscored the importance of these provisions in safeguarding the public from the risks associated with advanced AI systems. He found the resistance from OpenAI perplexing, given their previous commitments to safety assessments.
