Chris Lehane, OpenAI’s chief global affairs officer, confirmed that his company has been holding talks with Anthropic and Google DeepMind about AI safety for several weeks. He made the statement during a meeting with journalists on Tuesday, while Lehane is in Washington working with U.S. lawmakers to address catastrophic risks associated with AI.
The announcement confirms previous reports of undisclosed cooperation among the three companies and gives the discussion a practical dimension beyond general calls for safety. The article did not provide details about the mechanism of the talks or the decisions that resulted from them, and the source did not include comments from Anthropic, Google, or OpenAI in response to TechCrunch’s requests.
From a Call to Institutional Cooperation
The talks follow an article published by Dario Amodei, Anthropic’s CEO, in which he called on the AI sector to work together to slow the pace of developing advanced AI models and avoid catastrophic risks. The call received support from prominent figures, including Sam Altman of OpenAI, Demis Hassabis of Google, and Elon Musk of SpaceXAI.
Altman said OpenAI would join Anthropic in bringing in independent third-party evaluation organizations to monitor safety. Other reports also said that the three companies are working to establish an industry standards body, a step Altman told his employees might need to proceed without support from the U.S. government.
Hassabis had called in July for the creation of a new U.S. standards body that would serve as an AI oversight authority, with the power to inspect the most advanced models and coordinate a collective slowdown in development if risks escalated.
Coordination Faces a Competition Question
Cooperation among the largest AI laboratories is not without legal complexity. Some parties, including Altman, have noted that coordination among companies could raise risks related to antitrust laws if it were considered a means of limiting competition. Amodei proposed a limited government exemption allowing this type of cooperation on safety issues, but Lehane said the companies do not need such an exemption.
What Does This Mean in Practice?
The development represents a shift from requiring each company to evaluate its models internally toward exploring shared arrangements involving independent parties and industry standards. If these talks turn into a formal mechanism, they could affect how advanced models are tested before being released or their use is expanded. However, the source does not yet establish the existence of a joint body or binding agreement, nor does it explain how to ensure the independence of evaluators or prevent safety standards from being used to restrict competition.
At the same Tuesday meeting, Lehane announced OpenAI’s support for a provision in the FRONTIER Act requiring advanced AI laboratories to allow “independent verification organizations” to enter their companies to ensure that models are being developed safely.
This direction conflicts with the position of President Trump, who described safety concerns as a hoax and rejected the need for stricter regulation, arguing that any slowdown could give China an advantage. David Sacks, his AI adviser, echoed the view that concerns about AI’s existential risks are overstated. The next step therefore remains tied to the companies’ ability to turn voluntary cooperation into enforceable standards and to the position of U.S. lawmakers on the limits of this coordination.