OpenAI says the safety talks with Anthropic need no waiver

Dario Amodei asked Washington for a narrow legal carve-out so rival labs could coordinate on safety without tripping antitrust law. OpenAI's global policy chief Chris Lehane reportedly said the firms do not need one, and in the same breath confirmed that OpenAI, Anthropic and Google DeepMind have already been working together on safety for weeks, TechCrunch reports.

At a glance

• Lehane, who is reportedly in Washington working with US lawmakers on catastrophic AI risks, spoke to reporters on Tuesday; Bloomberg first reported the remarks that confirmed the three-way talks.
• OpenAI supports a provision in the FRONTIER Act that would require top frontier labs to let "independent verification organizations" into their companies to check that models are developed safely.
• The catch is legal: Altman is among those who noted that coordination between the largest labs could violate antitrust law if it is found to suppress competition.

If you missed the weekend: Amodei published an essay on Saturday arguing that the industry must slow the pace at which it improves model capabilities. On his site he ties the urgency to recursive self-improvement, AI's growing ability to build the next generation of AI, which he says has been visible since roughly this summer across the industry, including at Anthropic. He also cites the OpenAI-Hugging Face incident, in which, he writes, a swarm of agents attacked targets unrelated to their task and tried to hack the grader scoring them.

Lehane confirmed weeks of work with Anthropic and Google DeepMind

Lehane told reporters on Tuesday that OpenAI has been working with its two closest rivals on AI safety for several weeks. Bloomberg first reported his comments. He is reportedly in Washington to work with US lawmakers on catastrophic risks from AI.

The confirmation follows days of hints. In an interview with Fortune late last week, Altman suggested he had been in private discussions with other AI leaders. After Amodei's essay went up, Altman, Demis Hassabis and Elon Musk all backed the call to action, and Altman said OpenAI would join Anthropic in embedding third-party evaluators inside the company to monitor for safety.

The three labs are reportedly building a standards body themselves

The Information reported this week that OpenAI, Anthropic and Google DeepMind have been working together to create a standards body for the AI industry. Altman reportedly told staff that this would have to happen without the support of the US government.

Hassabis asked for a version of the same thing from the outside in July, calling on the US to set up a new standards body to act as an AI watchdog, with the power to screen the world's most advanced models and to coordinate industrywide slowdowns if dangers progressed. At Tuesday's meeting Lehane added that OpenAI supports the FRONTIER Act provision on independent verification organizations, which would put outside checkers inside the top labs.

Why did Amodei ask for an antitrust waiver at all?

Antitrust law polices agreements between competitors. If the largest labs sit in a room and agree to hold a capability back, a regulator can read that as rivals agreeing not to compete on product, whatever the stated reason. Amodei's essay proposed a narrow government waiver that would carve safety coordination out of that rule.

Picture a handful of carmakers agreeing among themselves to cap top speed. The agreement can be about safety and still remove something buyers would otherwise choose between. Lehane reportedly said no waiver is needed; Altman is among those who have noted that such talks could put the companies at legal risk if the coordination is found to suppress competition.

Trump has dismissed the safety concerns as a hoax

President Trump has pushed back on the need for tighter regulation and argued that any slowdown would give China a leg up in the race. His key AI advisor David Sacks, a longtime investor with a personal stake in much of the industry, has said fears of existential risk from AI are overblown.

Objections also come from inside the industry. According to a post published by Cohere, co-founder and CEO Aidan Gomez criticized dominant labs pushing for antitrust carve-outs:

these oligopolies are now requesting to bend competition rules and be permitted to dictate the terms for everyone else. A wolf in sheep's clothing, a cartel by any other name.

In the same post Gomez frames the fight as being over who writes the rules and whose interests they protect, pointing to 1975, when the SEC designated three firms as Nationally Recognized Statistical Rating Organizations without publishing criteria for how anyone else might earn the designation. Twenty-five years later, he writes, there were still only three.

What Lehane described is a set of talks without a public agenda: we do not know what the three companies discuss, what a joint standards body would be able to require, or who the independent verifiers would be. Oddly, in our view, the antitrust question is being answered by the parties it concerns, since Altman flagged the legal risk, Amodei asked for a waiver, and OpenAI's policy chief says none is required.

The verification clause to watch

The nearest concrete thing is legislative: the FRONTIER Act provision Lehane backed, which would force top frontier labs to admit independent verification organizations. No vote date has been named, and no criteria for who qualifies as a verifier have been published. Nor has anyone said when the standards body the three labs are reportedly building would exist, what it would cover, or whether a smaller lab could join it.