A U.S. government order shut down two of Anthropic's most powerful AI models, worldwide and without any transition period. What looks like a far-away news story touches on a risk every organization with AI applications faces: AI vendor lock-in. We lay out what happened and what it means for your AI strategy.

Friday, June 12, 2026, 5:21 PM ET.

Anthropic receives a letter from U.S. Secretary of Commerce Howard Lutnick. The order: immediately suspend all access to Claude Fable 5 and Mythos 5 for foreign users. Both models had launched barely three days earlier. Within hours, they were offline worldwide, for everyone, including European companies that relied on them daily.

Last week, this became reality.

What at first glance looks like a political conflict between the Trump administration and a Silicon Valley company touches on a fundamental risk for every organization using third-party AI models. It's the first time a leading tech company has been forced to take its most powerful AI models completely offline. It's probably not the last.

The underlying tensions between Anthropic and Washington have been simmering for a while. In February 2026, the Pentagon labeled Anthropic a supply chain risk, after the company refused to let go of two lines it wouldn't cross: no mass surveillance of American citizens and no fully autonomous weapons without human oversight. The direct trigger for last week's shutdown lay elsewhere. According to Anthropic, the Secretary's letter gave no concrete safety reason. The company inferred from verbal communication that the government was referring to a jailbreak technique, in which the model analyzes a codebase to find vulnerabilities. According to the Wall Street Journal, that vulnerability was reported to the U.S. Department of Commerce directly by Amazon.

Anthropic itself calls the measure a misunderstanding. The company states that the same capability is also available in other publicly deployable models, including GPT-5.5, and is working to restore access.

The order applied specifically to foreign users, including Anthropic's own foreign employees on American soil. Anthropic had no way to filter its users by nationality in real time. A worldwide shutdown was therefore the only way to comply with the order. Any European company building on these models was left out in the cold instantly. Without warning.

At Xylos, we see this pattern more often than you might think. Organizations build powerful AI applications, such as copilots, automation, knowledge management and customer interaction, all on top of a single external model. Speed of implementation takes priority, and the architectural risk often remains underexposed.

But what if that model is gone tomorrow? The cause doesn't have to be anything to do with you. A geopolitical conflict, a government decision, an acquisition, or a security incident is enough to suddenly cut off access.

This is what we call model dependency risk, and in our view it deserves a prominent place in every AI strategy.

We always advise our clients from the same principle: AI architecture should be vendor-agnostic and resilient by design.

Concretely, this means:

We remain convinced of the value of AI-as-a-service. The power of models like Claude and GPT is real, and what they deliver is substantial. Every powerful tool brings dependency with it, and that dependency deserves deliberate management.

Recent events show that the battle over who controls the future of autonomous intelligence is in full swing. Geopolitics and technology are today inextricably tangled up with regulation. As an organization, you can't influence that battle. But you can anticipate it.

Xylos helps companies build AI architectures that perform today and hold up against tomorrow's uncertainties. An architecture that's robust and agile, and that, above all, stays yours.

Do you have questions about how to make your AI strategy more resilient? We'd love to have that conversation.

About the author

Peter Verrykt is Business Unit Lead Data & AI at Xylos and helps organizations turn data into concrete business value. He helps companies look beyond technical implementations and uses data and AI as a foundation for better decisions, greater agility and sustainable growth.