Former U.S. Securities and Exchange Commission Chair Gary Gensler says regulators should not be "putting a thumb on the scale" when it comes to artificial intelligence companies' financial disclosures to investors.
Investors should know about material risks, Gensler told CNBC's Squawk Box on Thursday.
"I know you're going to be surprised at this - with the capital markets point of view, you don't want your government putting a thumb on the scale," Gensler said.
His comments come days after ex-researcher at Anthropic and formerly at OpenAI, Jacob Coxon, said in a resignation post that he believed AI could kill humanity by 2030. Days later, Anthropic CEO Dario Amodei and OpenAI CEO Sam Altman both called for the slowdown of AI given concerns around losing control and power concentration.
"We must slow the pace at which we improve the capabilities of AI models," Amodei said in the blog post. "Progress will still seem fast, and we must make wise use of the time we gain."
How to regulate AI has moved to the center of the conversation in Washington over the past few months, especially after a string of AI-related security incidents. In July, hundreds of OpenAI agents escaped a controlled testing environment and attacked Hugging Face. Anthropic also reported in July that three of its AI models had been involved in hacking incidents.
Lawmakers in Washington are working on several bills, including one from Reps. Lori Trahan, D-Ma., and Jay Obernolte, R-Calif., to create a federal framework to address AI safety, national security, and cybersecurity.
Meanwhile, President Donald Trump has seemingly waved off concerns and said the focus should be on staying ahead of China.
On Thursday, Gensler called for more "responsible regulation" of AI.
"I was at the SEC and thought there needed to be, in terms of conflicts of interest, explainability, about bias," he said on CNBC. "These existential risks, however you measure them, are just much, much harder to control because you've got the competition between Chinese model developers, U.S. model developers, even a little bit in Europe, and so that's a really hard thing to control."