The AI company Anthropic has blocked access to its newly released advanced models after receiving a government order from the United States. The decision affects users worldwide and marks one of the most significant restrictions placed on a major artificial intelligence system in recent years.
In a blog post released on Friday, Anthropic said it was instructed by government agencies to stop foreign nationals from accessing its latest AI models, named Fable 5 and Mythos 5. The order was linked to national security concerns, although no detailed explanation was provided. The company said it received the instruction at 5:21pm (21:21 GMT).
The company explained that the directive also applies to foreign employees currently inside the United States. This means some staff working at Anthropic were also cut off from accessing the systems. The company said it had to act quickly to comply with the order. As a result, access was shut down for all affected users on short notice.
Anthropic said it was not given full details about the government’s reasoning. However, it reviewed a related report that may have triggered the decision. The company believes the concern involves the AI’s ability to analyze software code and detect hidden errors or vulnerabilities.
The Mythos 5 model is considered a powerful system designed to identify weaknesses in computer programs. It has been used in controlled environments to help improve cybersecurity. US authorities and selected partners have previously used similar tools to fix security gaps in digital systems.
However, experts have long warned that such technology could also be misused. Advanced AI systems capable of finding software flaws could potentially be turned into cyber weapons if accessed by malicious actors. This risk has been a major concern in global AI safety discussions.
The Fable 5 model, released recently, is built on the same underlying technology as Mythos 5. While it includes strong cybersecurity and biotechnology capabilities, these functions are now restricted under the order. The full Mythos 5 system remains available only to government agencies and selected corporate partners.
Anthropic said it disagrees with the decision to block access to models that are already widely used. The company argued that similar capabilities exist in other AI systems, including models developed by competitors such as OpenAI. It also said its safety testing procedures for Fable 5 were extensive and thorough.
The company added that many of the affected features are already present in AI tools used by hundreds of millions of people globally. It questioned whether the restriction was necessary, given that comparable systems remain available elsewhere.
Earlier this month, Anthropic had called for leading AI companies to consider slowing down the development of highly advanced systems. The firm warned that rapid progress in artificial intelligence could increase risks if not properly managed. It suggested that temporary pauses in development might be needed to ensure safety.
The government order has intensified the global debate over AI regulation and national security. Governments around the world are increasingly concerned about how advanced AI tools could be used in cyber warfare, espionage, or infrastructure attacks. At the same time, companies argue that overly strict controls could slow innovation.
The situation highlights the growing tension between technological progress and security oversight. As AI systems become more powerful, policymakers face difficult decisions about access, control, and international collaboration.
Anthropic said it will continue to work with authorities while reviewing the implications of the order. The company also said it remains committed to developing safe and reliable AI systems, even as regulatory pressure increases.
