2 hours ago
2
An AI model so good at finding software vulnerabilities that the US government pulled the plug on foreign access. That’s the backdrop to French President Emmanuel Macron’s push at the G7 summit to restore allied nations’ ability to use Anthropic’s Mythos, a frontier AI system that has become both a prized cybersecurity tool and a geopolitical flashpoint.
On June 17, Macron said he expected progress “in weeks” on expanding non-US access to advanced American AI models. The statement came at the tail end of a three-day G7 summit where AI dominated conversations nearly as much as traditional economic coordination.
What happened and why it matters
Around June 13, Anthropic disabled global access to its two most advanced models, Mythos 5 and Fable 5, following what the company described as compliance with a US government directive citing national security concerns. No formal export control notification had been issued by mid-June. The restrictions instead arose from internal compliance measures Anthropic took in response to government pressure.
The reason is Mythos itself. By June 2026, the model had demonstrated the ability to autonomously identify thousands of zero-day vulnerabilities, the kind of undiscovered software flaws that intelligence agencies and cybercriminals alike would pay fortunes to exploit.
The G7 response: trust, but verify
The summit, held June 15-17, brought together not just heads of state but also the people building these systems. Anthropic CEO Dario Amodei and OpenAI’s Sam Altman were both present, a sign that AI governance has moved from academic panels to the highest levels of diplomatic engagement.
Macron met directly with Amodei during the summit. The central proposal on the table is what leaders are calling a “trusted partners” scheme, a framework that would allow controlled sharing of frontier AI capabilities with allied nations while maintaining safeguards against proliferation to adversaries.
The broader export control problem
The absence of a formal export control notification makes this even messier from a legal perspective. Anthropic acted on its own interpretation of government expectations, which means there’s no clear regulatory framework for other AI companies to follow.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.
English (US) · 