US Authorities Order Anthropic to Suspend Claude Fable 5 AI Model Over Security Concerns
Anthropic has halted the deployment of its Claude Fable 5 AI model following an order from US authorities. The company confirmed that foreign nationals are now barred from using the system, which it had previously described as “too powerful”.
“The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance,” Anthropic stated on its website. This development comes as Anthropic is already engaged in a legal dispute with the Trump administration over a directive preventing government agencies from utilising its AI tools.
US national security authorities reportedly believe a method for “jailbreaking” Fable 5 has been discovered. Anthropic acknowledged reviewing a demonstration that identified a limited number of “previously known, minor vulnerabilities.” The company maintains these vulnerabilities are relatively simple and can be exploited in other publicly available models without requiring such a bypass.
Ahead of its public launch, following a private release for testing in April, Anthropic had highlighted various “safeguards” incorporated into Fable 5 to prevent cyber exploitation. The company’s own assessment had deemed the tool capable of exploiting or hacking computer systems due to its advanced intelligence, leading to the self-proclaimed assertion that it was “too powerful to release.”
Professor Gina Neff of Queen Mary University London, an expert in Responsible AI, warned that such restrictions could impede the development and secure testing of advanced AI systems, potentially limiting international collaboration. Tests by the UK government’s AI Security Institute reportedly found the model capable of exploiting system defences 73% of the time.
The US Department of Defence previously labelled Anthropic a “supply chain risk”—a designation usually reserved for companies in adversarial nations—a decision Anthropic is currently challenging in court. A US judge has, for now, blocked the enforcement of this Pentagon directive.
