Anthropic has suspended access to its newly launched Claude Fable 5 and the more advanced Mythos 5 model after receiving an export control directive from the US government. The move comes only days after Fable 5 was introduced, making the restriction especially significant for customers, researchers, developers, and companies that were expecting to use the latest model. According to Anthropic, the directive requires the company to block access to both Fable 5 and Mythos 5 for all foreign nationals, regardless of where they are physically located.

This means the restriction does not only apply to users outside the United States. It also covers non-US individuals living or working inside the country, including foreign-national employees at Anthropic itself.

Because of the broad nature of the directive, Anthropic said it had to disable access to the two models for customers worldwide in order to remain compliant. The company added that its other AI models are not affected by the order.

A Major Restriction on Frontier AI Access

The decision marks one of the most direct examples of the US government stepping in to restrict access to advanced AI models on national security grounds.

AI export controls are not entirely new. Governments have already placed limits on certain advanced chips, high-performance computing technology, and AI-related infrastructure. However, restricting access to specific AI models is a much more visible and sensitive step.

Fable 5 and Mythos 5 appear to fall into what many describe as "frontier AI" territory, meaning they are among the more capable systems available or under development. These models can support a wide range of tasks, from software development and research assistance to advanced analysis and automation.

That flexibility is exactly why governments are paying closer attention. The same capabilities that make advanced AI useful for productivity and research can also raise concerns if they are misused.

US Officials Cite National Security Concerns

Anthropic said the US government did not provide full details about the specific national security concerns behind the order.

However, the company believes the directive may be connected to concerns that someone had found a way to bypass, or "jailbreak," Fable 5's safety controls.

In the AI world, a jailbreak generally refers to a method used to push a model outside its intended safety limits. These attempts may try to get a model to produce responses that it would normally refuse or redirect.

Anthropic disputes the seriousness of the reported issue. The company said it reviewed the demonstration and found that it only revealed a small number of previously known and relatively minor software vulnerabilities.

It also argued that other publicly available AI models could identify the same vulnerabilities without requiring any special jailbreak method.

Anthropic Says the Issue Has Been Misunderstood

Anthropic has described the government's directive as a misunderstanding and says it wants to continue working with US officials to resolve the issue.

The company appears to be taking the position that the risks linked to Fable 5 and Mythos 5 have been overstated, or at least interpreted too broadly. From Anthropic's perspective, the models were designed with safeguards, and the demonstration that triggered concern may not have shown anything uniquely dangerous.

Still, the company has complied with the order.

That is not surprising. When an export control directive is issued by the government, companies usually have very little room to ignore it, especially when the matter involves national security.

For Anthropic, the priority now is likely to clarify the situation, reassure regulators, and seek permission to restore access as soon as possible.

Why Foreign Nationals Are Affected Even Inside the US

One of the most striking parts of the directive is that it reportedly applies to foreign nationals regardless of location.

This means a person does not need to be outside the United States to be restricted from using the models. A foreign-national researcher, engineer, student, or employee inside the US could also be affected.

That may sound unusual to casual users, but it reflects how export control rules can sometimes work. In certain controlled technology areas, giving access to a non-US person can be treated as an export, even if the person is physically located inside the country.

For AI companies, this could create serious operational challenges. Global teams often include employees, contractors, researchers, and partners from many countries. If model access rules are applied by nationality rather than location, companies may need much stricter internal controls over who can view, test, or work with certain systems.

Fable 5 Had Just Been Released

The timing of the suspension is especially notable because Fable 5 had only launched earlier in the week.

Anthropic presented Fable 5 as its first Mythos-class model made broadly available to the public. It was positioned as a safer and more accessible version of Mythos 5, with additional safeguards designed to manage higher-risk requests.

According to the company, Fable 5 included systems that could redirect certain sensitive requests, including areas such as cybersecurity and biosecurity, to less capable models. The goal was to offer strong AI capabilities while still limiting access to functions considered more risky.

That makes the government's response particularly interesting. If Anthropic believed Fable 5 was the safer public-facing version, the directive suggests US officials may still have concerns about the model's overall capabilities or possible misuse.

Mythos 5 Also Restricted

While Fable 5 was the more public-facing model, Mythos 5 was described as the more advanced system.

It is not unusual for AI companies to maintain stronger internal or limited-access models while releasing more controlled versions to the public. These higher-capability systems may be used for research, testing, enterprise access, or future product development.

The fact that both Fable 5 and Mythos 5 are included in the directive suggests that the government is not only concerned about the public launch. It may also be looking at the broader model family and the level of access given to non-US individuals or organisations.

For Anthropic, this creates a larger issue than simply pausing one product rollout. It affects how the company manages access to its most advanced systems.

A Sign of Growing Government Scrutiny Over AI

This development fits into a wider global trend. Governments are increasingly trying to understand how to regulate advanced AI before the technology becomes even more powerful and harder to control.

For years, AI development moved faster than regulation. Companies launched new models, researchers tested capabilities, and businesses adopted AI tools quickly. But as models become more capable, governments are beginning to ask tougher questions.

These questions include:

The Anthropic directive may become an important example of how these questions are handled in practice.

Anthropic Previously Supported Stronger AI Oversight

The situation is also notable because Anthropic has previously argued in favour of stronger oversight for highly capable AI systems.

The company has publicly supported the idea of a global framework that could allow governments to temporarily pause the development or deployment of advanced AI if the risks are considered severe enough.

That position now has an interesting real-world test. Anthropic has advocated for oversight mechanisms, but it is now facing a government restriction that it believes may be based on a misunderstanding.

This highlights one of the biggest challenges in AI regulation. Oversight may be necessary, but if governments act too broadly or without clear technical explanation, companies may argue that innovation and access are being restricted unnecessarily.

The balance between safety, national security, and technological progress is becoming more difficult to manage.

What This Means for Users and Developers

For customers who had planned to use Fable 5 or Mythos 5, the immediate impact is simple: access is suspended for now.

Anthropic has said its other models remain available, so users may still be able to use alternative Claude models depending on their account, region, and eligibility. However, those who specifically needed the latest capabilities of Fable 5 or Mythos 5 will have to wait.

For developers and businesses, this could create uncertainty. Companies building workflows around frontier AI models may need to consider what happens if access changes suddenly due to regulation.

This is especially important for international teams. If advanced AI access can be restricted based on nationality or export control rules, businesses may need to rethink how they manage AI adoption across global operations.

The Bigger AI Policy Debate Is Just Beginning

The suspension of Fable 5 and Mythos 5 may not be an isolated event. It could be an early sign of how governments will treat the most advanced AI systems in the future.

As models become more capable, regulators may become more willing to intervene directly. That could include access restrictions, licensing requirements, safety evaluations, export controls, or mandatory reporting obligations.

AI companies will likely push for rules that are clear, technically informed, and consistent. Governments, meanwhile, will want to avoid situations where powerful AI tools could be misused in ways that affect national security.

The result may be a more controlled AI environment, especially for frontier models.

Final Thoughts

Anthropic's suspension of Fable 5 and Mythos 5 access shows how quickly the AI industry can move from product launch to regulatory crisis.

The company had only just introduced Fable 5 when the US government issued a directive restricting foreign access to both Fable 5 and Mythos 5. Anthropic says it disagrees with the basis of the concern and believes the issue has been misunderstood, but it has complied with the order while seeking a resolution.

For users, the suspension is a temporary disruption. For the AI industry, it may be a much bigger warning sign.

Advanced AI models are no longer being treated only as commercial software products. They are increasingly being viewed as strategic technologies with possible national security implications.

That means the future of AI access may depend not only on what companies build, but also on what governments decide should be controlled.