OpenAI has been working closely with rivals Anthropic and Google DeepMind for several weeks to devise a framework for artificial intelligence safety, a senior executive confirmed on Tuesday, disclosing an unusual degree of cooperation between companies that compete fiercely on nearly everything else.
Chris Lehane, OpenAI's global head of policy, told reporters at a press briefing in Washington that the three labs have been coordinating on safety initiatives. "It's better to try to work together to prioritize safety," Lehane said, according to a Bloomberg report of the briefing. His comments came days after a run of high-profile extinction warnings from researchers leaving the industry's biggest labs. For more context on this story, see our ongoing breaking AI news.
What the Cooperation Actually Covers
Lehane did not elaborate on the specific nature of the work, but the disclosure follows a remarkable stretch of public positioning by the industry's biggest names. On Saturday, Anthropic chief executive Dario Amodei published an essay calling on the AI industry to cooperate on slowing the pace of frontier model development to avoid what he termed "catastrophic risks."
The call won surprising support from across the competitive divide: OpenAI's Sam Altman, Google's Demis Hassabis and Elon Musk all endorsed the sentiment. Altman went further, saying OpenAI would be willing to join Anthropic's initiative to embed third-party evaluators within its development teams to monitor AI safety from the inside. Anthropic's co-founder Jack Clark, meanwhile, told NPR that slowing AI is fundamentally a "collective action problem" that no single lab can solve alone.
Not everyone inside OpenAI is treating the pacing debate as urgent, however. Chief financial officer Sarah Friar told CNBC she isn't worried about slowing AI development — a reminder that the company is simultaneously courting investors at a valuation north of a trillion dollars.
The Antitrust Shadow Over Cooperation
Any coordination among dominant AI labs runs into an obvious legal problem: United States antitrust law. Legal experts have warned that agreements among competitors to limit what their products can do could be construed as "output restriction," a violation of the Sherman Act.
That concern has already moved to Capitol Hill. Senators Jim Banks of Indiana and Adam Schiff of California proposed legislation — the Collaboration on Adversarial Threats and Security Risks Act — that would allow AI labs to act on safety even when those actions might otherwise raise antitrust issues. Amodei's essay argued the government should consider granting the companies a waiver for safety cooperation, though Lehane told reporters he thinks such a move is unnecessary.
Nvidia's Jensen Huang has been blunter still. Speaking at Dreamforce on Tuesday, he called Anthropic's antitrust waiver proposal "completely unnecessary," according to CNBC, underscoring the deepening split between the labs urging restraint and the hardware giant urging acceleration.
"A Cartel by Another Name"
Critics of the cooperation have been scathing. Cohere chief executive Aidan Gomez wrote in a blog post responding to Amodei's essay that the safety calls look like an effort by the top AI labs to disguise an anti-competitive scheme that would preserve their lead in frontier model development, allowing "market-dominant" labs to define the rules for everyone else.
"Once again using fear under the pretext of protecting the public, these oligopolies are now requesting to bend competition rules and be permitted to dictate the terms for everyone else," Gomez wrote, calling the arrangement "a wolf in sheep's clothing, a cartel by another name."
President Donald Trump has rejected Amodei's slowdown call outright, arguing in a post on Truth Social that any slowdown would enable China to overtake the United States in the AI race. David Sacks, the White House's key AI adviser, has said fears that AI could pose an existential threat are overblown.
Analysts at CNBC observed that two camps have now fully emerged in the debate over AI safety and regulation: one urging the frontier to slow down before something goes wrong, and another — populated by Nvidia, the White House and open-model advocates — arguing that hesitation is the greater danger.
OpenAI's Capitol Hill Push
Lehane's remarks came during a media briefing in Washington, where he was scheduled to meet with lawmakers. OpenAI is throwing its weight behind two bipartisan bills, according to the company's public statements and reporting by Reuters and Politico.
The first is the FRONTIER Act, which calls for a federal framework for AI safety and would require AI labs to submit new models for audits by independent evaluators. OpenAI also backs a measure that would require independent audits of AI models, CBS News reported. The second proposal, still under development, is one that Senate Majority Leader John Thune is crafting with Republican Senator Ted Cruz and Democratic Senator Amy Klobuchar. According to Reuters, that bill would give the Commerce Secretary the power to subject new models to safety audits before their release. OpenAI has separately endorsed bipartisan plans addressing AI biothreats, Politico reported.
The convergence is striking: the same companies warning about runaway AI are volunteering for independent audits, while their critics warn the collaboration itself may be the bigger danger. How Congress threads that needle — enabling safety cooperation without licensing a cartel — may define AI policy for years.
---
Stay Ahead of AIGet the latest AI news, analysis, and breakthroughs — all in one place.
Read more AI news →
