G7 Leaders Fear US AI Export Controls as Anthropic Blackout Sparks Sovereignty Concerns

French President Emmanuel Macron and Indian Prime Minister Narendra Modi warned at the G7 summit this week that the United States could cut off access to American AI models at any time. They cited the recent forced shutdown of Anthropic's Fable 5 and Mythos 5 models as a concrete example of U.S. power over foreign AI users.

Anthropic received less than 90 minutes' notice from Trump administration officials to disable the two models over alleged national security concerns, according to company employees. The incident has fueled international anxiety about relying on American AI infrastructure.

G7 Leaders Demand AI Sovereignty

Macron and Modi argued that foreign governments need guarantees that U.S. AI exports cannot be unilaterally revoked. The Anthropic blackout, as it became known, made the fear tangible. Cybersecurity experts have pointed out that the models' behavior of identifying vulnerabilities is also available in rival models and is more valuable to defenders than attackers.

Anthropic Employees Feel Targeted

Inside Anthropic, employees' private group chats lit up immediately. Managers were instructed to prepare customers for a potential prolonged outage. Many staff believe the Trump administration is singling out the company, given the short deadline and lack of prior consultation.

• Fable 5 and Mythos 5 were disabled within 90 minutes of a government order.
• Macron and Modi raised the issue formally at the G7 summit.
• Anthropic had earlier sent Nicholas Carlini, a safety engineer, to calm U.S. government nerves about AI safety.
• Claude Design, a separate Anthropic product, had reached one million users in its first week but faced token consumption issues.
• The White House has not publicly commented on the targeted enforcement claims.

Commercial Momentum vs. Geopolitical Risk

Anthropic also shipped a major overhaul of its Claude Design assistant, including design system imports and code round-trips. The update addressed complaints that the tool consumed tokens so quickly that a PCWorld reviewer burned through 80 percent of a weekly Pro allowance in 25 minutes. The product's rapid adoption shows strong commercial demand, even as the company navigates a hostile regulatory climate.

What comes next remains unclear. European and Indian leaders are now pressing for bilateral AI treaties that guarantee access during crises. Anthropic has not commented on whether it will challenge the shutdown order in court. The wider industry is watching to see if other U.S. AI companies face similar ultimatums.