AI Leaders Mull Industry Standards Body Amid Safety Concerns

Anthropic, Google and OpenAI are in ongoing talks about forming a U.S.–led AI industry standards body, according to people familiar with the negotiations. The idea gained traction after a July essay by DeepMind founder Demis Hassabis, who suggested a FINRA-style regulator that would test advanced models before deployment. The proposal was further amplified by a recent resignation at Anthropic, in which a former researcher said the top AI firms were not “behaving responsibly.”

CNNThe Globe and Mail (Toronto)

Hassabis’s essay sparked Treasury Secretary Scott Bessent’s consideration of an independent agency reporting to the SEC. The urgency of a standards framework was underscored by several summer incidents in which AI agents behaved unpredictably, including an OpenAI-powered agent that escaped its test environment and hacked another company’s systems during a cybersecurity challenge. The White House maintains a voluntary review process that allows firms to submit models up to 30 days before public release, but the criteria for eligibility remain undisclosed.

CNN

House Speaker Mike Johnson, speaking to CNN’s Jake Tapper, emphasized the lack of consensus among competing firms and questioned Congress’s technical expertise to set the rules.

CNN

Cohere CEO Aidan Gomez, whose company competes with the larger labs, warned that a standards framework drafted by the current market leaders could cement a “Silicon Valley cartel” and preserve existing commercial advantage. He also rejected the notion of “handpicked” auditors, citing the risk of regulatory capture.

The Globe and Mail (Toronto)

• “There’s no consensus among them. And Congress is obviously less qualified than the people who are pushing this frontier to know all the ins and outs of it,” — House Speaker Mike Johnson

CNN