OpenAI, Anthropic, and Google DeepMind have been discussing artificial intelligence safety for several weeks as leading AI companies explore ways to address risks associated with advanced frontier models.
OpenAI global policy chief Chris Lehane confirmed the discussions on Tuesday while speaking with reporters, following earlier reporting that the companies had been coordinating on AI safety efforts.
The talks come as the AI industry debates how companies should manage potential risks from increasingly capable models. Bloomberg reported that OpenAI has been working with Anthropic and Google on AI safety discussions.
AI companies discuss safety coordination
The discussions follow an essay by Anthropic CEO Dario Amodei calling for the AI industry to work together to reduce catastrophic risks from frontier AI development.
Several technology leaders responded publicly to Amodei’s call, including OpenAI CEO Sam Altman, Google DeepMind CEO Demis Hassabis and SpaceX CEO Elon Musk. Their responses included support for stronger safety measures and industry cooperation.
Altman said OpenAI would work with Anthropic on embedding third-party evaluators to monitor safety, while Hassabis also supported efforts around AI safety standards. Musk also weighed in on the broader discussion.
Companies weigh standards and regulatory questions
The discussions have raised questions about how competing AI companies can cooperate on safety without limiting competition. Some industry leaders have warned that coordination between major companies could create antitrust concerns.
Amodei proposed a limited government waiver that would allow AI companies to coordinate on safety issues, but Lehane said the companies do not believe such a waiver is necessary.
The talks also follow reports that AI companies have been exploring the creation of an industry standards body. The Information reported on efforts to create a framework for AI oversight, while Altman has previously discussed the need for broader industry coordination.
Altman also discussed AI safety issues in a recent interview with Fortune, including concerns around controlling advanced AI systems.
Political debate continues over AI safeguards
The discussions are taking place alongside a wider debate in Washington over AI regulation. Google DeepMind’s Hassabis has previously called for a new standards body that could evaluate advanced AI models and coordinate responses to potential risks.
The Biden administration has not been the only voice in the debate. President Donald Trump has pushed back against calls for tighter AI restrictions, arguing that slowing development could weaken the United States’ position against China. Reuters reported on Trump’s comments regarding AI regulation and data centres.
Other public discussions around the issue have continued among policymakers and technology leaders, including Hassabis’ previous comments on AI oversight and public coverage of the ongoing AI safety debate.
OpenAI supports independent model evaluation.s
During the same meeting, Lehane said OpenAI supports a provision in the FRONTIER Act that would require leading AI labs to allow independent verification organisations to assess whether models are being developed safely.
The proposal reflects a broader shift toward external evaluation as governments and AI companies consider how to monitor increasingly capable systems.
