OpenAI, Anthropic, and Google DeepMind have been engaged in private discussions for several weeks focused on AI safety, according to Chris Lehane, OpenAI’s global policy chief. These talks come as AI leaders in the industry seek to address the potential catastrophic risks linked to advanced artificial intelligence. Lehane revealed the collaboration while in Washington, where he is meeting with lawmakers to discuss safety frameworks. The conversation follows an influential essay by Anthropic CEO Dario Amodei calling for a collective industry effort to slow AI’s development pace to manage risks.

Notable figures like OpenAI’s Sam Altman, Google’s Demis Hassabis, and SpaceXAI’s Elon Musk publicly supported Amodei’s proposal, emphasizing the importance of embedding third-party evaluators into AI companies to monitor safety practices. However, concerns about antitrust laws have surfaced, as some experts warn that such coordination could potentially reduce competition between AI labs. Despite this, Lehane stated that the companies do not believe they require a government waiver to pursue safety cooperation, countering Amodei’s suggested narrow legal exemption.

Reports also indicate that the three companies are exploring the formation of a standards body to oversee AI development and safety protocols, a move Altman has said must occur independently of U.S. government support. This follows earlier calls from Hassabis for a regulatory body that would vet cutting-edge AI models and coordinate industry-wide pacing efforts. Meanwhile, former President Trump and his AI advisors have rejected the urgency of AI safety concerns, fearing that regulation could hinder U.S. competitiveness against China.

Lehane expressed OpenAI’s support for provisions in the bipartisan FRONTIER Act, which would mandate leading AI labs to permit independent safety assessments by external organizations. As the conversation around AI safety escalates, these talks reflect the growing recognition within the industry that coordinated measures and oversight are critical to managing the risks posed by increasingly powerful AI technologies. The outcomes of these discussions may shape future regulatory and safety standards for AI development worldwide.