Anthropic has recently closed its most advanced artificial intelligence models from global access with a sudden decision taken by the US government. On June 12, the Trump administration made it mandatory for Fable 5 and Mythos 5 models to be used only by US citizens.

This restriction turned into a worldwide access problem because Anthropic did not have sufficient infrastructure to verify its user base. The company was forced to halt all global access, including to researchers in allied countries and its own foreign-born employees, with 90 minutes’ notice.

NSA Theses and Security Concerns

According to The Economist, the information shared by Senator Mark Warner sheds light on the underlying reason behind this sudden ban. General Joshua Rudd, leader of NSA and US Cyber ​​​​Command, stated that the Mythos model provided access to nearly all of NSA’s classified systems within hours during an authorized penetration test.

The argument in question shows that although the model was developed for cybersecurity research, it can bypass even the most protected infrastructures. This situation stands out as the most important reason that explains why the government intervened so quickly and harshly.

When Anthropic announced the Mythos model in April, it stated that it would not release a public version due to the model’s high ability to detect security vulnerabilities. The model was being tested in a limited way with 200 select partners such as Amazon, Google, Microsoft and Nvidia within the scope of Project Glasswing.

Fable 5 was released to the public on June 9 and was released as a version with added security layers. Government officials argue that these security layers are not sufficient to prevent the cyber attack capacity of the model.

Political Tension and Future Uncertainty

The official trigger for the ban was the argument that Fable 5 was a way to bypass security systems. This situation, reported by Amazon to the Ministry of Commerce, increased concerns that Anthropic’s defense mechanisms were bypassable.

Anthropic objected to the decision, stating that this vulnerability was narrow in scope and that similar situations could occur in other artificial intelligence models. The company warned that if these standards are applied to the entire segment, new model distributions may stop completely.

The intelligence agencies of the Five Eyes countries published a joint statement that artificial intelligence models will radically change both offensive and defensive capacities. This shows that cyber risks are no longer just a technical nuisance but a core business and leadership responsibility.

For now, Fable 5 and Mythos 5 models remain closed for many users. Anthropic maintains its expectation that the models will be available again soon.

How do you think such harsh export restrictions on artificial intelligence models will affect the future of the technology world?