Trump reverses on Anthropic threat label after G7 talks, signals possible rollback of export controls

From threat to handshake

President Donald Trump said in a pre-taped Axios interview published on Friday that he no longer views Anthropic as a national security threat. Asked whether he considered the AI company a threat, Trump replied, “Well, not now, but a week ago, maybe.” He described meeting CEO Dario Amodei at the G7 summit in Évian-les-Bains on Wednesday, walking away with the impression that Amodei was “nice” and “smart.”

“He responded to us very quickly because you know it’s a tremendous liability. People get put in prison immediately for that. You can’t play games with that. And he responded very responsibly, I thought.”

Trump signalled he may ease the restrictions on Anthropic’s Fable 5 and Mythos 5 models, saying, “I would, but I’m not sure I have to do that,” when asked about a potential rollback.

A week of crisis

On June 12, the Commerce Department ordered Anthropic to seek US government approval before foreign nationals could access its most powerful AI systems, Fable 5 and Mythos 5. The directive followed months of escalating tension. In March, the Pentagon designated Anthropic a supply-chain risk after the company refused to strip safety guardrails related to surveillance and autonomous weapons from products used by the US military. Commerce Secretary Howard Lutnick subsequently sent a letter threatening criminal charges. The situation intensified when Amazon CEO Andy Jassy alerted the administration about an alleged vulnerability in Fable 5. The White House imposed export controls, and Anthropic suspended global access to both models within roughly 90 minutes.

The carve-out that survived

The shutdown was not total. Organizations that had been granted early access to a more restricted version, Mythos Preview, under the Glasswing programme retained it. Industrial cybersecurity firm Dragos and Cisco Systems both confirmed they still had access. The European cybersecurity agency ENISA, however, was told it would no longer be given access, reversing an earlier invitation. The episode highlighted the degree of discretion Anthropic holds in interpreting the export directive.

“We are grateful to the administration for their ongoing partnership in working to get this matter resolved as quickly as possible. We remain committed to working alongside them towards our shared goals of protecting critical infrastructure and making sure the U.S. leads in AI.”

Allies push back

The ban drew sharp reactions at the G7. French President Emmanuel Macron warned, “We won’t buy any models made by these companies if overnight, you can just flip the switch.” Indian Prime Minister Narendra Modi called for broad and inclusive access to US AI models, while the UK requested a carve-out that was rejected. A group of more than 170 tech executives signed an open letter saying the curbs risked America’s AI leadership.

The IPO shadow

Anthropic had confidentially filed for an initial public offering in early June, with a valuation Fortune reported at approximately $965 billion. The federal restrictions had cast uncertainty over the listing. Trump’s conciliatory tone, and the possibility of de-escalation, could stabilise investor confidence ahead of the offering.