Anthropic has suspended access to two of its newest artificial intelligence models after U.S. authorities raised national security concerns, escalating a dispute between the company and the Trump administration over the use of advanced AI.
The San Francisco-based firm said it had been ordered to restrict foreign nationals from using Claude Fable 5, a model released only days earlier after a limited private testing period. The order also affected Mythos 5, prompting Anthropic to disable both systems more widely for customers.
“The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance,” the company wrote.
The company said officials had not identified a specific flaw in the model, but had raised concerns about a possible way to bypass its safety controls. Such methods, often called jailbreaking, are used to push AI systems beyond their intended limits or make them ignore restrictions built into the software.
“Our understanding is that the government believes it has become aware of a method of bypassing, or ‘jailbreaking’ Fable 5,” Anthropic said.
Anthropic said it reviewed the demonstration cited by authorities and found it had been used to detect a small number of known, minor software weaknesses. The company argued that other publicly available models could find similar flaws without any bypass being required.
Before its release, Anthropic had promoted Fable 5 as its most capable general model yet, while also stressing that extra safeguards had been added to reduce cyber security risks. The company’s earlier description of the system as unusually powerful drew criticism from some observers, who saw it as partly promotional.
The suspension has renewed questions about how governments should oversee frontier AI systems without limiting the research needed to test them safely. In Europe, officials said the case underscored the importance of reducing reliance on foreign technology providers.
Anthropic is also challenging a Pentagon decision to label the company a “supply chain risk.” A U.S. judge has blocked enforcement of that directive while the case continues.