Anthropic has suspended its latest AI models, Claude Fable 5 and Mythos 5, just days after their public release, following security concerns raised by US authorities.
In a statement, the company said it was ordered to prevent foreign nationals from using Claude Fable 5, a tool Anthropic itself described as "too powerful." To comply, Anthropic abruptly disabled both models for all customers.
While US national security authorities have not specified their exact concerns, Anthropic believes the government became aware of a method to "jailbreak" Fable 5—bypassing software restrictions to access sensitive information or unlock features. However, the company noted that the vulnerabilities identified were minor and could also be found in other publicly available models without needing a bypass.
Claude Fable 5 is an iteration of Anthropic's Claude Mythos, an AI program that competes with OpenAI's ChatGPT and Google's Gemini. Prior to release, Anthropic touted safeguards against hacking, but concerns from finance, tech, and government leaders persisted. The company had granted pre-release access to select organizations due to the model's advanced capabilities, which it claimed could be dangerous if exploited.
The suspension adds to ongoing tensions between Anthropic and the Trump administration. The US Defense Secretary previously labeled Anthropic a "supply chain risk"—a designation typically reserved for foreign adversaries. Anthropic is suing the Pentagon over the matter, and a judge has temporarily blocked its enforcement, allowing government agencies to continue using Anthropic's tools while the lawsuit proceeds.
Anthropic emphasized that Fable 5's capabilities exceed any model it has previously released, but critics have questioned whether the "too powerful" narrative was overstated for marketing purposes.