When Fable and Mythos disappeared

On Friday Anthropic switched off access to its two most powerful models, Fable 5 and Mythos 5, following a directive from the US government. The justification was export control and national security: no foreign nationals are to have access, wherever they are.

As far as I know, this is the first time intelligence itself has been export-controlled. Not the chips, not the tools around it: the model itself. And this is only the beginning. Perhaps the model comes back soon, but Pandora’s box has been opened.

The point is simple: states will treat frontier AI as a strategic national resource, on a par with ITAR, and they will not let private companies hold independent control. On Friday we got the first concrete example, less than a week after launch.

For Europe this should accelerate both the conversation and the plans for strategic autonomy. When access can disappear overnight on another nation’s terms, autonomy is no longer an academic exercise. It is a capability discussion: what do we actually own in the stack? ASML is strategic infrastructure. European alternatives are not a luxury, but increasingly a critical necessity.

We have little time, and we are poorly coordinated. Both are urgent.