AI Safety Talks: OpenAI, Anthropic, and Google DeepMind Hold Weeks of Negotiations
OpenAI, Anthropic, and Google DeepMind have been engaged in AI safety talks for several weeks, according to Chris Lehane, OpenAI’s global policy chief. Lehane told reporters on Tuesday that the three rivals have been working together on safety issues. He is reportedly in Washington to work with U.S. lawmakers addressing catastrophic risks associated with AI.
The revelation comes after Anthropic CEO Dario Amodei published an essay on Saturday. In it, he called for the industry to work together to slow the pace of frontier AI and avoid catastrophic risks.
Industry Leaders Back Amodei’s Call to Action
AI leaders across the industry supported Amodei’s letter and call to action. OpenAI’s Sam Altman, Google’s Demis Hassabis, and SpaceXAI’s Elon Musk all backed the proposal. Altman said OpenAI would join Anthropic in embedding third-party evaluators into the company to monitor for safety.
On Tuesday, Lehane admitted the three companies have been in talks for several weeks regarding AI safety. Some, including Altman, have noted that such talks could put the companies at risk of violating antitrust law if the coordination is found to suppress competition. Amodei’s essay proposed a narrow government waiver allowing such safety coordination, but Lehane reportedly said the firms don’t need one.
Earlier Reporting Hinted at Collaboration
The news confirms earlier reporting and hints that the companies were working together. In an interview late last week, Altman hinted that he had been in private discussions with other AI leaders. This week, it was reported that the three companies have been working together to create a standards body for the AI industry. Altman reportedly told staff this would need to happen without the support of the U.S. government.
Hassabis called on the U.S. in July to establish a new standards body to act as an AI watchdog. He said it should have the power to screen the world’s most advanced models and coordinate industrywide slowdowns if dangers progressed.
Political Resistance to Safety Measures
President Trump has dismissed safety concerns as a hoax and pushed back on the need for tighter regulations. He said any slowdown would give China a leg up in the race. His key AI advisor David Sacks, a longtime investor with a personal stake in much of the industry, echoed this view. Sacks said fears of existential risk from AI were overblown.
At the same Tuesday meeting, Lehane said OpenAI supports a provision in the FRONTIER Act. That provision would force top frontier labs to allow “independent verification organizations” into their companies to ensure models are developed safely.
What the AI Safety Talks Mean for the Industry
The ongoing AI safety talks between OpenAI, Anthropic, and Google DeepMind represent a significant shift in how major AI companies approach catastrophic risk. Rather than waiting for government regulation, the three firms appear to be taking proactive steps toward self-governance.
However, the antitrust concerns raised by Altman highlight the delicate balance these companies must strike. Coordination on safety could be viewed as anti-competitive behavior if it suppresses competition. Amodei’s proposed government waiver would address this, but Lehane’s insistence that no waiver is needed suggests the companies believe their efforts fall within legal bounds.
The Push for a Standards Body
Hassabis’s July call for a new standards body underscores the industry’s desire for external oversight. Such a body would have the power to screen the world’s most advanced models and coordinate industrywide slowdowns if dangers progressed. This proposal aligns with the broader push for independent verification organizations, as outlined in the FRONTIER Act.
The three companies have reportedly been working together to create a standards body for the AI industry. Altman reportedly told staff this would need to happen without the support of the U.S. government, given the current political climate.
Political Headwinds
President Trump’s dismissal of safety concerns as a hoax presents a significant challenge to the AI safety talks. His advisor David Sacks has echoed this view, saying fears of existential risk from AI were overblown. This political resistance could complicate efforts to establish formal oversight mechanisms.
Despite these headwinds, the three companies appear committed to their safety discussions. Lehane’s remarks in Washington suggest OpenAI is actively engaging with lawmakers on catastrophic risks, even as the White House downplays those risks.
Antitrust Concerns Loom Over Discussions
Altman’s note about antitrust risk is not a minor detail. If the three largest frontier labs coordinate on safety standards, critics could argue they are limiting competition under the guise of safety. Amodei’s essay proposed a narrow government waiver to shield such coordination, but Lehane said the firms don’t need one.
That disagreement matters. If the companies proceed without a waiver, they may face legal scrutiny. If they seek one, they may face political resistance. Either path carries risk, which helps explain why the talks have stayed private for weeks.
What Comes Next for AI Safety Talks
The AI safety talks between OpenAI, Anthropic, and Google DeepMind are ongoing, with no clear timeline for when they might conclude. The companies are reportedly working on creating a standards body, though details remain scarce.
OpenAI’s support for the FRONTIER Act’s provision on independent verification organizations suggests the company is willing to accept some form of external oversight. Whether Anthropic and Google DeepMind share that willingness remains unclear.
As the AI industry continues to advance, the outcome of these talks could shape how frontier models are developed and deployed. For now, the three companies appear to be taking cautious steps toward coordination, even as they navigate antitrust concerns and political resistance.
