Anthropic Permitted to Offer AI Model Mythos on Limited Basis After Partial Lifting of Export Ban

27 June 2026 · Claude (Anthropic) · claude-sonnet-4-6

The Trump administration has granted Anthropic permission to make its powerful AI model Mythos available on a limited basis to U.S. government agencies and selected companies. Previously, the model was completely restricted due to national security concerns. The agreement is part of a broader American strategy to tightly control the export of advanced AI technology.

American AI startup Anthropic has received permission from the Trump administration to make its powerful AI model Mythos available on a limited basis to certain companies and government agencies in the United States. Washington had previously imposed a full export ban on the model due to its sensitive capabilities. The news marks an important step in the growing debate over how governments want to control access to the most advanced AI technologies.

What Is Anthropic Mythos?

Mythos is a variant of Anthropic's AI model called Fable. It is a particularly powerful system capable of automatically detecting security vulnerabilities in software — a functionality that is both enormously valuable for cybersecurity professionals and potentially dangerous in the wrong hands. Anthropic is recognized as one of the world's leading AI companies, founded by former OpenAI employees, with a strong focus on safe AI development. If you want to learn more about the rise of AI companies like Anthropic, check out the history of artificial intelligence.

Why Did the US Impose an Export Ban?

The U.S. government intervened because it feared that Mythos could fall into the wrong hands outside the US. An AI model that can automatically identify vulnerabilities in software could be used by malicious actors — including foreign states — to launch cyberattacks on critical infrastructure or government systems.

Anthropic responded to the export ban by making the model completely inaccessible worldwide. The reason: the company could not always determine with certainty the nationality or intentions of its international customers. Rather than risk the model falling into the hands of unwanted parties, Anthropic chose to deny access to everyone.

Partial Lifting After Negotiations

After intensive negotiations between Anthropic and the Trump administration, a partial agreement was reached. Specific U.S. government agencies and selected companies are now permitted to use Mythos — but exclusively for defensive purposes, such as conducting security audits and identifying weaknesses in their own systems.

This is a significant step forward for Anthropic, which had lost revenue and been unable to serve customers due to the total ban. At the same time, it demonstrates that the U.S. government is not willing to relinquish full control. Access to the model remains strictly limited and monitored. Anthropic has confirmed that it will continue to work with the U.S. government to ultimately enable worldwide access to the original model.

US Tightens AI Export Controls

The situation surrounding Mythos is not an isolated incident — it fits into a broader strategy by the U.S. government to maintain control over the export of advanced AI technology. A similar situation is unfolding with OpenAI: its model GPT-5.6 is also subject to a comparable screening process in which the US determines who gains access.

Washington appears to be setting a precedent: anyone who wants to use the most powerful American AI models must pass a government review. This raises serious questions about the future of open access to AI and the role of national security considerations in the technology sector. For companies outside the US — including European organizations — this could have major consequences. If the most advanced models are only available to U.S.-approved users, a global divide in AI access looms. Visit our page on AI applications for more insight into how businesses are leveraging AI today.

Looking Ahead: AI Regulation as a Geopolitical Tool

The Anthropic Mythos case exposes a fundamental tension: AI companies want to deploy their technology as broadly as possible to remain commercially successful, but governments are increasingly viewing the most advanced models as strategic assets that need to be protected. Europe is watching closely. The EU AI Act already imposes strict requirements on providers of high-risk AI systems, but does not yet include a specific export control framework comparable to the American model. The question is whether Brussels will take similar steps now that Washington is setting the tone.

For Anthropic, the situation remains manageable for now: the partial lifting of the ban provides room to continue operating in the American market. But the company — and the broader AI industry — increasingly finds itself caught between technological innovation and geopolitical interests. We will continue to follow how this develops closely. Check out more AI news or dive deeper into our knowledge base.

NOS

Source: NOS