Anthropic Blocks Access to Claude Fable 5 and Mythos 5 Models at US Government Request

Anthropic has complied with a US government request to block access to its latest AI models, Claude Fable 5 and Claude Mythos 5. The decision affects all global users, including corporate clients and internal teams. The move stems from undisclosed national security concerns, though Anthropic labeled the situation a "misunderstanding" and pledged to restore access as soon as possible.

The government’s intervention may have been triggered by a public breach of the Fable 5 model, reported by a researcher under the pseudonym Pliny the Liberator. The researcher claimed to have bypassed the model’s safeguards, gaining access to instructions for creating cyberweapons, explosives, and chemical compounds. However, Anthropic noted that the evidence provided by the government was unconvincing, as similar vulnerabilities exist in other public models, including OpenAI’s GPT-5.5.

Experts warn that this incident exposes the regulatory risks of centralized AI models. Enterprises are advised to diversify their AI sources by leveraging multiple providers or local solutions to avoid disruptions in critical operations. Some companies are already shifting to open-source models that can be deployed on their own infrastructure, reducing reliance on external factors.