The United States is drawing a new line around artificial intelligence. Anthropic must block access to its top models Fable 5 and Mythos 5 for foreign nationals — even if they live in the US or work for Anthropic. Because the company cannot implement the order safely in any other way, the models are being disabled for all customers for now.
The United States is now visibly treating advanced artificial intelligence as a security-critical key technology. By order of the US government, Anthropic must block access to its models Fable 5 and Mythos 5 for foreign nationals. The rule applies not only to users outside the United States, but also to foreign nationals inside the country — including Anthropic’s own employees.
Anthropic is responding drastically. Because the requirement is technically and legally difficult to limit cleanly to individual user groups, the company is disabling Fable 5 and Mythos 5 for all customers for now. Other Anthropic models are expected to remain available.
The move is unprecedented. Until now, US export controls in the AI sector have mainly targeted chips, high-performance computers, data centers and technical infrastructure. Now the focus is no longer only on hardware, but on the AI model itself. This shifts the debate: advanced AI is no longer being treated merely as a software product, but as a potential instrument of power.
An intervention with signal effect
The US government is justifying the order on national security grounds. Specific details were not initially made fully public. The main concern appears to be that the models could be misused through so-called jailbreaks — for example to analyze software vulnerabilities or support sophisticated cyberattacks.
Mythos 5 in particular is considered especially powerful. The model was originally intended only for selected partners. Fable 5 was presented as a more strongly secured version for broader use. Only a few days after launch, the government-mandated shutdown has now followed.
For the AI industry, this is a turning point. A government is no longer merely regulating the general framework, but restricting access to concrete AI systems. This affects customers, developers, companies and even the provider’s own employees.
A new category of AI control is emerging. Access to the most powerful models will no longer be decided only by pricing, contract terms or technical activation. Citizenship, export law and national security interests may become the decisive boundary.
AI becomes geopolitical
The case shows how far the AI debate has shifted in a short period of time. Only a few years ago, generative AI was mainly about text, images, productivity and automation. Today, it is about cyber capabilities, military usability, strategic dependencies and technological competition with China.
The US wants to prevent especially powerful AI systems from falling into the wrong hands. But the term “wrong hands” is becoming politically charged. If foreign nationals are broadly excluded from access, a new digital border is created — right through the global AI community.
This is especially explosive because AI research itself is international. Many developers, researchers and specialists in the United States are not US citizens. If export controls affect not only companies abroad, but also people inside American labs, the way the industry works could change dramatically.
For Anthropic, the situation is delicate. The company has positioned itself strongly around safety, control and responsible AI. Now this very safety debate is becoming a problem. Its own models are apparently considered so sensitive that the state is restricting access to them.
The myth grows through the ban
Paradoxically, the shutdown could make Fable 5 and Mythos 5 even more famous. Even before the order, reports of allegedly spectacular capabilities of the new models were circulating on social media. Programmers in particular shared examples claiming that Fable 5 had restructured complex codebases, identified security issues and generated large amounts of code.
Such reports should be treated with caution. An impressive social media post is not a reliable performance benchmark. Especially in programming, AI can generate code that looks professional but does not actually work. That is one of the central risks of advanced models: they can produce convincing results without being reliable.
Still, the combination of user hype and government restriction is powerful. A model that is blocked for security reasons shortly after launch automatically gains aura. When the state limits access, many people hear it as an unintended seal of quality: this AI must be exceptionally strong.
That is why some observers describe the situation as a perfect myth machine. Whether intended or not, the ban makes the models more desirable. It creates scarcity, mystery and the feeling that something has emerged that goes beyond ordinary product logic.
Between security risk and marketing effect
The case also shows the difficult conflict facing AI companies. They have to prove to investors, customers and the public that their models are becoming more powerful. At the same time, they have to prove that this power remains controllable.
The stronger a model appears, the greater its market value becomes. But that very strength brings regulators, security agencies and militaries into the picture. AI companies are therefore moving through an increasingly narrow corridor: they sell progress, but must also prove that this progress will not become dangerous.
Anthropic appears to dispute the government’s assessment at least in part. The company reportedly sees the measure as excessive and points out that the underlying risks are limited or also present in other models. Nevertheless, in the short term, it has little choice but to comply with the order.
For customers, this is a warning signal. Anyone who bases business-critical processes on specific AI models must now reckon not only with technical outages or price changes. Political interventions, export rules and security decisions can also abruptly change access.
A new phase of AI control
The shutdown of Fable 5 and Mythos 5 therefore marks more than a dispute between Anthropic and the US government. It reveals a new reality: the most powerful AI systems are increasingly being treated as strategic technologies.
This has consequences for companies, developers, governments and users worldwide. AI will not simply remain globally available just because it can be delivered over the internet. The more powerful models become, the more strongly governments will try to control access to them.
A new phase is beginning. After the debate over chips and data centers, the models themselves are now moving to the center of geopolitical control. The question is no longer only who can build AI. The question is also who is allowed to use it.
Anthropic has shut down Fable 5 and Mythos 5. But the real story is bigger: advanced AI has finally arrived in the realm of national security.
SK
