OpenAI, Anthropic, Google have been in talks on AI safety for weeks

Chris Lehane, OpenAI’s planetary argumentation chief, told reporters connected Tuesday that the institution has been moving with rivals Anthropic and Google DeepMind connected AI information for weeks, arsenic archetypal reported by Bloomberg. Lehane is reportedly successful Washington to enactment with U.S. lawmakers addressing catastrophic risks associated with AI.

The revelation comes successful the aftermath of Anthropic CEO Dario Amodei’s essay, published Saturday, which called for the manufacture to enactment unneurotic to dilatory the gait of frontier AI and debar catastrophic risks. AI leaders crossed the industry, including OpenAI’s Sam Altman, Google’s Demis Hassabis, and SpaceXAI’s Elon Musk, supported Amodei’s missive and telephone to action, with Altman saying that OpenAI would articulation Anthropologic successful embedding 3rd enactment evaluators into the institution to show for safety. 

On Tuesday, Lehane admitted that the 3 companies person been successful talks for respective weeks regarding AI safety. Some, including Altman, person noted that specified talks could enactment the companies astatine hazard of violating antitrust instrumentality if the coordination is recovered to suppress competition. Amodei’s effort projected a constrictive authorities waiver allowing specified information coordination, but Lehane reportedly said the firms don’t request one. 

The quality confirms earlier reporting and hints that the companies were moving together. In an interrogation with Fortune precocious past week, Altman hinted that helium had been successful backstage discussions with different AI leaders. This week, The Information reported that the 3 companies person been moving unneurotic to make a standards assemblage for the AI manufacture — thing Altman reportedly told unit would request to hap without the enactment of the U.S. government. 

Hassabis called connected the U.S. successful July to found a new standards body to enactment arsenic an AI watchdog, with the powerfulness to surface the world’s astir precocious models and coordinate industry-wide slowdowns if dangers progressed. 

President Trump, for his part, has dismissed information concerns arsenic a hoax and pushed backmost connected the request for tighter regulations, saying immoderate slowdown would springiness China a limb up successful the race. His cardinal AI advisor David Sacks — a longtime capitalist with a personal stake successful overmuch of the manufacture — echoed this, saying fears of existential hazard from AI were overblown.

At the aforesaid Tuesday meeting, Lehane said that OpenAI supports a proviso successful the FRONTIER Act that would unit apical frontier labs to let “independent verification organizations” into their companies to guarantee models are developed safely.

TechCrunch has reached retired to Anthropic, Google, and OpenAI for comment.