Tech CEOs pitch a U.S.-led AI coalition at G7

Just days after the White House yanked the cord on Anthropic's latest models over national security concerns, the company's CEO Dario Amodei was sitting at lunch with Donald Trump in France.

At a closed-door lunch during the G7 summit in Évian-les-Bains, Amodei teamed up with Google DeepMind CEO Demis Hassabis to pitch a U.S.-led global coalition to write the rulebook for frontier AI.

The guest list was stacked. Trump was flanked by Treasury Secretary Scott Bessent, Commerce Secretary Howard Lutnick, and Secretary of State Marco Rubio. OpenAI chief Sam Altman was also at the table alongside about a dozen tech executives and world leaders.

Drawing the battle lines

Amodei didn't mince words about what this coalition should actually do. According to attendees, he called for international rules governing structured access to frontier models, alongside a chip and hardware trade ring that explicitly shuts out China.

He also pushed for allied governments to pool resources to counter AI threats in cyber warfare, bioterrorism, and intelligence operations.

The pitch found an immediate fan in Canadian Prime Minister Mark Carney, who agreed the U.S. should take the wheel. OpenAI global affairs chief Chris Lehane later noted that other foreign leaders in the room also acknowledged Washington's natural role as standard-setter.

OpenAI's Altman pitched his own angle, proposing an international forum to hammer out globally accepted evaluation benchmarks and deliver independent risk assessments.

Politics over compute

The timing of this gathering is hard to ignore. Last Friday, Anthropic was forced to shut off access to its newest flagship models, Fable 5 and Mythos 5, after sudden U.S. export controls hit the firm. The company is currently stuck in high-stakes negotiations with the Trump administration.

Meanwhile, OpenAI has already started playing the national security card, rolling out a specialized GPT-5.5 Cyber model last month in a heavily guarded preview for vetted cybersecurity pros.

Why it matters

If you build software for a living, take note of the phrase structured access. When labs start calling for government-backed access frameworks, it usually means the days of simply punching in a credit card to get unfiltered frontier API access are ending for the most capable models.

For the broader industry, this closed-door lunch looks like the formal merger of frontier AI research and Western defense policy. By actively asking the Trump administration to lead a cartel that controls chips and model distribution, labs like Anthropic and OpenAI are essentially trading away operational independence in exchange for a government-sanctioned moat.

We think Amodei's move is clever damage control. Anthropic just took a massive hit from Washington's export controls, so volunteering to help design the allied rulebook is the best way to get Fable and Mythos back online.

The obvious risk here is that when national security and AI licensing merge, smaller startups and open-source researchers will inevitably get locked out of the frontier.