The world's most powerful AI shut down after 48 hours: when Washington decides a model is too dangerous

There are moments when technology stops being a product and becomes a matter of State. That is precisely what happened between June 10 and June 12, 2026, when Anthropic publicly released Fable 5, the consumer-facing interface of its internal Mythos 5 system, and the United States government ordered, within forty-eight hours, that access be disabled for all foreign users, regardless of the country in which they were located.

Those who managed to use Fable 5 during those two days describe experiences that are difficult to frame within the parameters to which artificial intelligence had accustomed us. Tasks that normally required full days of work were completed in a matter of hours. Software projects that with other frontier models would have demanded dozens, or even hundreds, of hours were compressed into radically shorter time windows. These were not incremental improvements. Those who used it speak of a perceptible qualitative leap, almost physically unsettling in its effectiveness.

Then, on June 12, 2026, came the shutdown.

The official justification, communicated by American authorities, was national security. According to information that emerged, internal research conducted by Amazon — one of Anthropic's principal investors through an agreement worth billions of dollars — had identified a method to bypass certain protections in the model, exploiting it to detect advanced cybersecurity vulnerabilities. This was not a trivial exploit: what was being described was the capacity to turn a commercial AI assistant into a tool potentially useful for cyberwarfare operations or for hostile actors seeking to map critical infrastructure.

Anthropic responded in an unexpected fashion, publicly contesting the government's decision. The company argued that the identified problem was limited in scope and not universal, and pointed out that analogous capabilities already existed in other frontier models currently available on the global market. A technically sound argument, but a politically slippery one: many observers noted that Anthropic had been among the most vocal AI companies in calling for sector-wide regulation, and that it now found itself challenging the exercise of that very same regulatory authority.

But beyond the specific dispute, what happened marks something far larger. For years, the American strategy of technological containment had focused on chips. The Export Administration Regulations, updated multiple times between 2022 and 2025, had built a formidable barrier around Nvidia's most advanced GPUs, preventing their export to China and other countries deemed strategic adversaries. This was control over hardware, over silicon, over the physical substrate of artificial intelligence.

What happened with Fable 5 is different. Here the subject is controlling software, the model, the trained system itself. It is as if, after years of restricting the sale of centrifuges, someone had decided that nuclear physics manuals also fall within dual-use technologies. Artificial intelligence is being treated, for the first time in an explicit and operational manner, on a par with enriched uranium, reconnaissance satellites, or advanced missile systems. This is a paradigm shift that redefines the technological geopolitics of the coming decades.

And then there is the other story, the human one, perhaps more unsettling. Fable 5 existed as a public product for less than forty-eight hours. Yet within that extraordinarily brief window of time, millions of people around the world had already begun to consider it indispensable. Technical forums, social networks, and developer communities were already filled with workflows built around Mythos 5, professional pipelines that incorporated it, and accounts from people who had already restructured the way they worked.

We adapted to the most powerful artificial intelligence ever released to the public in less time than it typically takes to finish a television series. This says something about how structurally dependent we have already become on these tools, and about how little time it takes for a technology to move from novelty to perceived entitlement. The real problem, perhaps, is not that the American government shut down Fable 5. The real problem is how quickly we stopped imagining the world without it.

Sources: official statements from the United States Department of Commerce; Anthropic public statement dated June 12, 2026; Amazon-Anthropic investment agreement as documented by Reuters and The Verge.