Two former researchers from OpenAI have raised urgent concerns regarding the company’s opposition to California’s proposed SB 1047, a bill designed to enforce strict safety protocols in AI development, including the implementation of a “kill switch.”
In a letter addressed to California Governor Gavin Newsom and first shared with Politico, William Saunders and Daniel Kokotajlo expressed their disappointment at OpenAI’s stance, which they found unsurprising given their experiences with the company.
The researchers, who initially joined OpenAI with the intention of ensuring the safety of its powerful AI systems, stated, “We resigned from OpenAI because we lost trust that it would safely, honestly, and responsibly develop its AI systems.”
They emphasised that developing advanced AI models without sufficient safety measures poses significant risks of catastrophic harm to the public.
Despite OpenAI CEO Sam Altman’s public advocacy for AI regulation, the researchers pointed out a contradiction in his actions. They noted that while Altman has called for government intervention in AI oversight during congressional testimonies, he opposes specific regulatory measures when they are proposed. This inconsistency has led to scepticism about the company’s commitment to safety.
An OpenAI spokesperson responded to these criticisms, asserting that the company’s position on SB 1047 has been mischaracterised. They referred to a letter from Chief Strategy Officer Jason Kwon addressed to California Senator Scott Wiener, the bill’s sponsor, which outlined the company’s reasons for opposing the legislation.
In Kwon’s letter, he acknowledged that SB 1047 “has inspired thoughtful debate” and indicated that OpenAI supports some of its safety provisions. However, he argued that due to national security implications, AI regulation should be developed at the federal level rather than through state legislation.
Kwon stated, “A federally-driven set of AI policies, rather than a patchwork of state laws, will foster innovation and position the US to lead the development of global standards.”
This perspective has been met with scepticism by Saunders and Kokotajlo, who argue that the push for federal legislation is not the only reason for OpenAI’s opposition. They criticised the company’s objections to SB 1047 as “not constructive and don’t seem in good faith.”
The former researchers expressed urgency in their call for action, stating, “We cannot wait for Congress to act—they’ve explicitly said that they aren’t willing to pass meaningful AI regulation.” They highlighted the risk that federal legislation, if it ever materialises, could preempt California’s efforts to establish necessary safety measures.
Ultimately, Saunders and Kokotajlo urged the California Legislature and Governor Newsom to pass SB 1047 into law, expressing hope that with appropriate regulations, OpenAI could still fulfil its mission of developing AGI safely.