In a surprising shift toward cooperation among the world’s biggest tech rivals, OpenAI, Anthropic, and Google DeepMind have spent the last several weeks holding private talks focused on AI safety. Chris Lehane, OpenAI’s global policy chief, confirmed the collaboration during a trip to Washington where he has been engaging with lawmakers about the potential for catastrophic risks tied to advanced artificial intelligence. This move follows a public plea from Anthropic CEO Dario Amodei, who recently argued that the industry must collectively slow its pace of development to prevent existential disasters.
The alliance has gained significant momentum with backing from heavyweights like Sam Altman and Demis Hassabis. As part of this effort, OpenAI has signaled it is open to bringing in third party evaluators to monitor its internal processes for safety flaws. However, this level of coordination brings its own set of legal headaches. Some executives, including Altman, have voiced concerns that collaborating too closely with competitors could trigger antitrust investigations if regulators believe these meetings are being used to stifle competition rather than protect humanity. While Amodei suggested seeking a government waiver to legalize these interactions, Lehane indicated that such formal permission might not be necessary.
Beyond immediate safety protocols, there are reports that the trio is attempting to build an independent standards body for the entire AI sector. This initiative appears to be moving forward privately, potentially bypassing official government channels despite previous calls from Hassabis for a state backed watchdog with the power to halt dangerous model releases. Such efforts face a steep uphill battle in the current political climate; President Trump has previously dismissed many AI safety warnings as a hoax, arguing that strict regulations would only serve to hand a strategic advantage to China.
Despite the political friction, OpenAI continues to push for legislative guardrails through proposals like the FRONTIER Act. Lehane stated that his company supports mandates that would require top laboratories to grant access to independent verification organizations. By inviting outside eyes into their proprietary systems, these companies hope to prove they are managing risk responsibly even as they race toward more powerful versions of general intelligence. For now, the industry remains caught between a desire for self regulation and a volatile regulatory environment in Washington.