What happens now to Fable 5 and Mythos 5?

On 12 June 2026 the US government, citing national security authorities, issued an export control directive that requires Anthropic® to “suspend all access to Fable™ 5 and Mythos™ 5 by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees.” The stated reason: what the directive frames as a “method of bypassing, or ‘jailbreaking’ Fable 5,” specifically the act of “asking the model to read a specific codebase and fix any software flaws.” In other words, exactly what every security engineer does every day.

Source: Anthropic: Statement on the US government directive to suspend access to Fable 5 and Mythos 5 →

Anthropic itself publicly disagrees: “we disagree that the finding of a narrow potential jailbreak should be cause for recalling a commercial model deployed to hundreds of millions of people.” If this bar were applied across the industry, every new model deployment would grind to a halt.

When watts and amps become a compliance incident

I can speak from recent experience about where this logic leads. Not long ago I asked Fable 5 how to size the power supply for a workstation with three GPUs. The answer: refused, on the grounds that I was apparently in violation of policy because I clearly wanted to “build a better AI”. A question every electrician answers in two minutes turns into a compliance incident.

When a model can no longer talk about watts and amps out of fear of misuse, that is not a safety gain, it is a productivity loss. And now it disappears entirely. Not because it was too dangerous, but because a political process moved faster than a technical one.

Safety measure or political retaliation?

Is this a measure grounded in safety, or is it payback for the recent public dispute between the US administration and Anthropic? From the outside, that cannot be answered with certainty. But three observations sit in the room: the timing right after the dispute, Anthropic’s own characterisation as “the finding of a narrow potential jailbreak,” and its own observation that “other publicly-available models are able to discover them as well.” Taken together, the comfortable safety justification gets thin. Were export control law to become leverage against inconvenient providers, that would be an industrial policy problem, not a safety question.

A risk to every business model built on US AI

That brings us to the core point for every entrepreneur. Anyone whose business model leans on a single US model, whether in product development, customer support, automation, or a SaaS product, can be cut off overnight without warning. Not because of a bug, not because of a missed payment, but with the stroke of a pen out of Washington. Create dependence, and use it as leverage when convenient. Whether that was the plan here, no outsider can say. Whether the pattern fits a strategy is for each reader to judge. For anyone in Europe building on AI as a productive force, this is a wake-up call. Supplier risk is no longer just a matter of price and SLA, it is a matter of geopolitics.

Have we Europeans been asleep?

While the US and China have been pouring billions into foundation models, GPU capacity and data centres for years, we have busied ourselves with AI Acts, ethics committees and funding programmes for “European champions” that mostly stayed at the prototype stage. By a rough estimate our own models currently sit one to two years behind the leading US and Chinese foundation models. Mistral is a good start, but too small, too underfunded, and on its own not capable of carrying an entire industrial ecosystem. What we have is regulation. And regulation does not power up a single server.

The real question

So the real question is not whether Fable 5 will come back. It is: what consequences do we draw from this for our architectures, our contracts, our second-source strategies, and for our industrial policy stance as a continent? And who makes the next “safety” call, on what factual basis, and with what motive?

Keywords: Fable 5, Mythos 5, Anthropic, US export control AI, Fable 5 review, AI vendor risk, European AI sovereignty, Mistral, multi-vendor AI architecture consulting, LLM second-source strategy, AI consulting Switzerland