In a recent correspondence, Jason Kwon, the Chief Strategy Officer at OpenAI, strongly advocates for federal government intervention in AI regulations. As previously reported by Bloomberg, Kwon argues that an AI safety bill being considered in California could impede progress and potentially lead companies to relocate out of the state. Kwon asserts that a comprehensive set of AI policies driven by the federal government, as opposed to a patchwork of state laws, will promote innovation and position the United States as a global leader in AI development. Consequently, OpenAI, along with other AI labs, developers, experts, and members of California's Congressional delegation, respectfully opposes SB 1047 and welcomes the chance to express key concerns. The intended recipient of the letter is California State Senator Scott Wiener, who initially introduced SB 1047, also known as the Safe and Secure Innovation for Frontier Artificial Intelligence Models Act. According to proponents like Wiener, the bill establishes standards prior to the advancement of more powerful AI models.

It mandates precautions like pre-deployment safety testing and other safeguards, provides whistleblower protections for employees of AI labs, grants the California Attorney General the authority to pursue legal action in the event of harm caused by AI models, and proposes the creation of a "public cloud computer cluster" named CalCompute. In response to the published letter, Wiener highlights that the proposed requirements apply to any company conducting business in California, regardless of their state of origin, rendering the argument against the bill illogical. Wiener also notes that OpenAI does not critique any specific provision of the bill and concludes by stating, "SB 1047 is an extremely reasonable bill that simply asks large AI labs to fulfill their existing commitments, which involve testing their large models for catastrophic safety risks. " After receiving concerns from figures like Zoe Lofgren and Nancy Pelosi, support from companies like Anthropic, and backing from organizations such as California's Chamber of Commerce, the bill underwent various amendments during the committee stage. These adjustments include replacing criminal penalties for perjury with civil penalties and narrowing the pre-harm enforcement capabilities of the Attorney General. The bill is currently awaiting final voting before it can be presented to Governor Gavin Newsom for approval. Please find below the complete text of OpenAI's letter.