Advanced AI Models Mythos 5 and Fable 5 Are Available Again

The US government has lifted the access restriction placed on Anthropic's Mythos 5 and Fable 5 AI models. Following the decision by the US Department of Commerce, the company will be able to offer its customers the models it recently had to shut off.

In a statement on its official X (formerly Twitter) account, Anthropic announced that it had obtained the necessary permissions from the Department of Commerce and would begin gradually reopening access to the Mythos 5 and Fable 5 models as of July 1, 2026.

What Were the Reasons Behind the Ban?

About two weeks ago, in line with national security and technology protection policies, the US government imposed an obligation on Anthropic to block foreign nationals from accessing its new AI models.

This rather harsh regulation covered not only users abroad but also foreign-national staff working at Anthropic within US borders. Not wanting to face legal sanctions and aiming for full compliance, Anthropic decided to fully suspend global access to the Mythos 5 and Fable 5 models after this restriction.

The suspension came right after cybersecurity concerns were raised about the autonomous capabilities Mythos-class models displayed in penetration tests and claims that they could break through NSA systems within hours.

A Temporary Exception for Critical Infrastructure and Project Glasswing

While the official restrictions were in place, the US government granted a limited exception for Claude Mythos 5 ahead of its final decision. Under this special arrangement, the model was made available to certain organizations that operate and protect critical infrastructure in the US (energy, communications, finance and so on).

Around 100 organizations within the Project Glasswing partnership continued to have access to Anthropic's most advanced cybersecurity-focused AI model during this period. However, the future of the Claude Fable 5 model, which aims to bring Mythos's superior coding and reasoning capabilities to a general audience, remained uncertain.

Howard Lutnick: "Security Risks Will Be Managed Proactively"

After weeks of intensive negotiations, the matter was settled. US Secretary of Commerce Howard Lutnick shared the details of the agreement reached with Anthropic.

Lutnick announced that Anthropic agreed to the following commitments:

* Proactively identify and address the security risks associated with the models, * Work rigorously with the US government on protocols, standards and safe releases for Mythos, Fable and future advanced models, * Immediately report to the US government any malicious or suspicious activity attempted through the models.

New Safety Measures and Classifiers from Anthropic

In its official statement, Anthropic said it had stepped up safety measures for the redeployment of the models:

> "Claude Fable 5 will be available again globally starting tomorrow. Following productive discussions with the US government, we are re-releasing the model with a new set of classifiers designed to target and block cybersecurity-focused misuse."

The company also acknowledged that these strict new safety classifiers might initially affect some routine software development tasks as well:

> "In the near term, some routine tasks such as coding may get caught by the cybersecurity filters and be blocked by mistake. We are continuing our work to optimize these filters and deliver a smooth experience for our developers."

Agentic Capabilities Are Being Released in a Controlled Way

The lifting of the restrictions on the Mythos 5 and Fable 5 models shows that the development of autonomous agents in the AI industry will continue without interruption. However, this agreement with the US Department of Commerce is also clear proof of how critical a role government oversight and national security protocols will play in the development of future "frontier" AI models.