News

U.S. Restricts Access to Anthropic’s Fable 5 and Mythos 5 AI Models Over Security Concerns

• By John K. Waters
• 06/12/2026

Anthropic has restricted access to its newest artificial intelligence models, Fable 5 and Mythos 5, after the U.S. government ordered limits on foreign access due to national security concerns, according to the company and people familiar with the matter.

The Commerce Department directive requires Anthropic to prevent foreign nationals from accessing the models, citing concerns that advanced systems could be used to identify software vulnerabilities or to support cyber operations by foreign intelligence services. Anthropic has said it complied by broadly restricting access while it assesses how to implement narrower controls.

The models, released earlier this month, represent the company's most capable systems to date. Anthropic described Fable 5 as a general-purpose model with additional safety filtering, while Mythos 5 was positioned as a higher-capability variant intended for vetted users in cybersecurity and infrastructure roles.

U.S. officials have raised concerns that the systems could be “jailbroken” or otherwise misused to help identify weaknesses in widely used software. Anthropic has disputed the characterization of the risk, arguing that similar capabilities exist in other publicly available AI systems and that the issue has been overstated.

The order marks one of the most significant uses of U.S. export-control-style authority for a frontier AI system, according to legal experts and industry analysts. It also highlights growing tension between AI developers and regulators over how advanced models should be governed, particularly when they are delivered through globally accessible cloud-based services.

Anthropic executives met with U.S. officials in Washington this week to seek clarification on the scope of the restrictions and to argue for more targeted safeguards. People familiar with the discussions said the company is seeking a framework that allows continued international access under stricter controls, rather than broad geographic limitations.

Amazon, a major investor in Anthropic and a provider of its cloud infrastructure, was among the technology firms that had previously raised concerns with government officials about the potential cybersecurity implications of advanced AI models, according to a source familiar with the matter. Amazon declined to comment.

The dispute has drawn attention across the technology industry, where companies are increasingly weighing the risks of regulatory intervention against the benefits of rapid model deployment. Some cybersecurity experts have supported tighter oversight, while others have warned that restricting access to a single provider’s systems may do little to reduce overall risk, given the availability of competing models, including open-source alternatives.

Legal analysts say the case could become a reference point for how U.S. export controls are applied to software-based AI systems. Unlike traditional controlled technologies, frontier AI models are typically accessed remotely, raising questions about how nationality-based restrictions should be enforced in practice.

The Commerce Department has not publicly detailed the full legal basis for the order. Officials have previously indicated that advanced AI systems could fall under existing authorities designed to regulate technologies with potential military or intelligence applications.

Anthropic has said it is working to restore access where possible while complying with government requirements. The company has not said when full functionality for international users might return.

The outcome of the dispute is expected to influence broader policy discussions in Washington over whether new legislation is needed to more explicitly regulate frontier AI systems, or whether existing export-control frameworks are sufficient.