G7 leaders: Stop U.S. from cutting off Anthropic AI access

G7 leaders warned the United States could revoke access to top American AI models, arguing countries cannot build critical tools on technology that Washington can "turn off the switch" from one day to the next. The comments came at the G7 Summit on June 18, where leaders pressed tech executives and each other over the risks of U.S. export controls.

What happened at the G7 summit?

At the G7 Summit on June 18, leaders including French President Emmanuel Macron and Indian Prime Minister Narendra Modi raised concerns about unilateral U.S. control of AI access, and they discussed a potential "trusted partners" scheme to keep advanced models available to allies. Macron made his remarks during a lunch that included top AI executives such as Anthropic CEO Dario Amodei and OpenAI CEO Sam Altman, and he warned that if the U.S. "from one day to the next can turn off the switch," it would hurt customers and the firms themselves.

Delegates debated how to preserve access to leading American models while addressing security and safety complaints. The meeting featured conversations about granting access to non-U.S. nations and companies designated as trusted partners, with the explicit aim of using those models to build defenses against rivals like China.

How did the U.S. block of Anthropic’s models happen?

The Trump administration blocked Anthropic from exporting its newest Mythos 5 and Fable 5 models, an action taken after Amazon flagged to the White House that certain safety guardrails on those models could be bypassed. The article notes this sequence as the immediate trigger and adds that cybersecurity experts argue the cited capabilities are also present in models that remain publicly available, including offerings from OpenAI.

The export block has left Anthropic’s models unavailable internationally, exposing a structural risk: companies and governments that rely on U.S. infrastructure can lose access overnight for reasons they may never be fully told. Prime Minister Modi expressed concern over the decision, a point cited in reporting from the Financial Times, and Macron suggested Washington should back broader access, including to Mythos, so customers would not buy U.S. AI that could vanish without warning.

Why it matters

The episode sharpens the problem of technological dependence. If a few U.S. firms supply the most capable models, allied nations and companies face both economic disruption and strategic vulnerability when access can be revoked. Industry voices amplified that worry: Cohere co-founder and CEO Aidan Gomez said dependence on a small handful of big tech companies is dangerous to resilience, and framed digital sovereignty as more than market competition.

The dispute also highlights a policy tension: national security officials may seek to restrict exports of models they deem risky, while partner governments want predictable, unfettered access to build critical infrastructure and defenses. That tension could reshape procurement choices, investment in local models, and international AI governance.

What to watch

Watch whether Washington formally backs a "trusted partners" scheme that would allow vetted non-U.S. nations and companies to access advanced models from firms like Anthropic and OpenAI, and whether Mythos access is reopened more broadly. Another signal will be whether other major cloud or AI vendors raise similar export concerns to U.S. authorities, which could trigger further restrictions or wider diplomatic pushback.