US blocks Anthropic's Fable 5 and Mythos 5 AI models, triggering European sovereignty alarm

The US move

The US government ordered Anthropic to block foreign nationals from accessing its latest frontier models, Claude Fable 5 and Mythos 5, citing national security. Anthropic said it was told to restrict access on Friday and could not technically implement the order in time, so on Saturday it took both systems offline globally.

“Government agencies directed us to block access for all non-US persons under reference to national security.”

Fable 5 had been released only days earlier as a public version of the more powerful Mythos 5, which was already limited to selected partners because of its advanced capabilities in finding software vulnerabilities and supporting complex programming tasks.

European alarm

German security and digital policy officials reacted sharply. Marc Henrichmann, chair of the Bundestag's intelligence oversight committee, called the move a security dependency.

“When a single administrative act in Washington decides overnight which AI tools are available for European authorities, companies and researchers, that is not a technical detail, but a security-political dependency.”

Konstantin von Notz, vice chair of the Green parliamentary group, said the decision was extremely short-sighted and predicted massive negative effects on demand for US providers. Interior minister Alexander Dobrindt warned that Europe must rapidly develop its own AI capabilities to avoid becoming a victim in an AI arms race.

Economic and IPU fallout

The block also threatens Anthropic's rumoured public listing. Harrison Rolfes of US research firm Pitchbook noted that if a government can de facto disable a company's core product, its valuation at IPU could plunge.

“If the government can switch off leading AI models, how can a company go public? It could lose a huge amount of value practically immediately.”

Anthropic's rapid rise had already drawn scrutiny, and the Trump administration had previously clashed with the firm over military access to its systems.

Calls for digital independence

The EU Commission, which had been in talks with Anthropic for weeks to secure access for the bloc's cybersecurity agency ENISA, said those conversations continue. A spokesperson stressed that technology is increasingly a strategic value and Europe must be able to act according to its own principles.

German officials and media commentators alike used the incident to demand a European digital "Apollo programme" – building own AI models, computing infrastructure, and data centres to reduce dependency on US providers.

Unclear motives

While the official reason is national security, the exact threat has not been made public. Anthropic speaks of a misunderstanding. US media have reported that Amazon, a rival and investor in Anthropic, may have tipped off the government. The episode adds to tensions between the Trump administration and a company that resisted giving the military access to its AI.