California’s AI security bill could potentially slow down innovation?

The CEO of OpenAI, Jason Kwon, maintains in a newly penned letter that guidelines on artificial intelligence should be formulated and implemented by federal authorities. According to previous reports, Kwon notes that a proposed AI security bill in California could potentially slow down innovation and prompt companies to relocate out of the state if not carefully considered.

To accelerate AI innovation nationwide, a unified federal framework of AI insurance policies would be more effective than the current disparate state laws, enabling the U.S. To navigate the evolution of global demands. As a result, we join various AI labs, builders, specialists, and members of California’s Congressional delegation in respectfully opposing SB 1047, and we appreciate the opportunity to highlight several crucial concerns that have guided our stance.

The letter is addressed to California State Senator Scott Wiener, referring to the “Secure and Safe Innovation for Frontier Synthetic Intelligence Fashion Act.”

Accordingly, the proposed legislation builds upon Wiener’s advocacy by stipulating prerequisites in anticipation of exceptionally powerful AI systems, mandating pre-deployment security assessments and other protective measures, while also shielding whistleblowers in AI research facilities, empowering California’s Attorney General to take legal action should AI models cause harm.

Accordingly, Wiener concluded on Wednesday night that the proposed requirements pertain to any business operating in California, regardless of headquarters location; thereby rendering the argument unnecessary. OpenAI does not contest a single provision of the bill, and it is noted that SB 1047 merely demands what large AI labs have already agreed to do: scrutinize their massive models for catastrophic security threats.

The invoice awaits final approval before being sent to California Governor Gavin Newsom’s office for signature.

OpenAI, a non-profit artificial intelligence research organization, today announced that it will no longer sell its language model technology to companies that use it for military purposes. The decision comes after widespread criticism of the organization’s initial stance on the issue, which was seen as overly permissive by some.