OpenAI, Anthropic And Google DeepMind Have Been In AI Safety Talks For Weeks

OpenAI has confirmed that it has been discussing artificial intelligence safety with rivals Anthropic and Google DeepMind for several weeks, bringing previously private coordination between three leading frontier AI developers into public view, according to Reuters.

The talks come amid growing calls from AI leaders for stronger independent model evaluations and shared safety standards, while the involvement of direct competitors has also raised questions about how such coordination fits within antitrust rules.

 

TL;DR

• OpenAI says it has held AI safety talks with Anthropic and Google DeepMind for several weeks.
• OpenAI policy chief Chris Lehane says the companies do not need an antitrust waiver simply to coordinate on safety.
• The discussions follow calls for independent AI evaluations and common frontier-model safety standards.
• No formal three-company safety alliance or finalized standards body has been announced.

 

OpenAI Confirms AI Safety Talks With Anthropic And Google DeepMind

 

OpenAI Global Policy Chief Chris Lehane confirmed the discussions during a briefing in Washington, according to Reuters, citing Bloomberg News.

Lehane said OpenAI had been speaking with Anthropic and Google DeepMind for several weeks about AI safety and did not believe an antitrust exemption was required for the companies to coordinate on such matters.

"It's better to try to work together to prioritize safety," Lehane said, according to Reuters.

However, this does not mean OpenAI, Anthropic and Google DeepMind have agreed to jointly slow development, established binding safety rules or created a formal AI safety organization.

TechCrunch reported that the discussions have taken place as the companies explore how the frontier AI industry could improve safety practices, including through independent verification.

 

Why Are Rival AI Companies Talking About Safety?

 

The disclosure follows a renewed push from Anthropic CEO Dario Amodei for AI developers to give safety measures more time to keep pace with increasingly capable frontier models.

In his essay, Dario Amodei called for stronger external evaluation and greater coordination between developers.

He also argued that the U.S. government would "need to issue a narrow waiver for certain kinds of safety conversations" between competing AI companies.

OpenAI, however, appears to disagree that such an exemption is necessary for the safety discussions currently taking place, according to Reuters, as syndicated by WTVB.

Google DeepMind CEO Demis Hassabis had previously proposed creating an independent standards organization capable of evaluating advanced AI systems and establishing common expectations for frontier-model safety, according to TechCrunch.

Furthermore, reports suggest OpenAI, Anthropic and Google DeepMind have discussed the possibility of an industry standards body, although no finalized organization has been publicly announced.

Independent AI Model Evaluations Gain Support

 

One area where the companies appear to be moving closer together is independent evaluation.

Lehane said OpenAI supports provisions that would allow independent verification organizations to assess whether frontier AI systems are being developed safely, according to TechCrunch.

This also follows a wider shift toward external testing of advanced AI systems.

Reuters recently reported that U.S. lawmakers have been discussing requirements for companies developing powerful AI systems to work with outside or government experts on model testing and verification.

OpenAI has separately expressed support for mandatory national AI safety requirements.

 

Are There Antitrust Concerns?

 

Safety collaboration between competitors comes with another complication: antitrust law.

Anthropic's proposal for a narrow exemption drew scrutiny from U.S. Federal Trade Commission Chairman Andrew Ferguson.

Speaking in his personal capacity, Ferguson said regulators should be cautious when incumbent companies simultaneously seek new regulations and exemptions from competition rules, according to Reuters, as syndicated by WTVB.

Lehane's position is different.

OpenAI does not believe an antitrust waiver is required merely for OpenAI, Anthropic and Google DeepMind to communicate about AI safety, according to Reuters.

The distinction could become important if discussions move beyond sharing safety practices and evaluations toward coordinated decisions affecting when or how competing companies release advanced models.

For now, the confirmed development is narrower: three of the companies competing at the frontier of generative AI have been talking about how to manage safety risks together.

Whether those discussions eventually produce common testing requirements, an independent standards organization or another formal framework remains to be seen.