OpenAI, Anthropic and Google DeepMind are coordinating on AI Safety.

OpenAI has been discussing AI safety with rivals Anthropic and Google DeepMind for several weeks, a rare attempt by the three frontier-model developers to coordinate as concern grows over the pace and risks of increasingly capable AI systems.

OpenAI global policy chief Chris Lehane disclosed the talks on September 15 while meeting policymakers in Washington. He said cooperation was preferable when addressing safety and argued that the companies should not need a special antitrust waiver to work together on the issue. No formal agreement between the three labs has been announced, and the precise scope of their cooperation remains unclear. Separate reporting indicates the companies have discussed creating an industry body that could develop common safety standards. OpenAI CEO Sam Altman has also backed allowing independent evaluators inside frontier AI companies to assess advanced systems.

The conversations predate the unusually public display of agreement among AI executives over the weekend. Anthropic CEO Dario Amodei called for the industry to slow the pace of frontier-model development to give safety measures more time to catch up. Altman, Google DeepMind CEO Demis Hassabis and SpaceXAI’s Elon Musk subsequently expressed support for major elements of his proposal.

Antitrust law has quickly become part of the debate. Amodei proposed a narrow legal exemption allowing AI companies to coordinate development pacing without being accused of suppressing competition. Lehane said OpenAI does not believe such an exemption is necessary. US Federal Trade Commission chair Andrew Ferguson separately warned against broad antitrust carve-outs, while legal experts have argued existing law already permits cooperation designed to prevent genuine catastrophic risks. OpenAI is also backing bipartisan US legislation targeting catastrophic AI risks, including provisions for independent verification of frontier systems.

The development does not amount to an industry-wide pause. Instead, it signals that companies competing intensely on models, talent and computing power are increasingly considering common safeguards, even as regulators debate how much coordination between those same competitors should be allowed.