White House Forces Anthropic to Disable Claude Fable 5 and Mythos 5

Anthropic, the embattled AI company behind the Claude line, has been forced to take its most advanced models—Claude Fable 5 and Claude Mythos 5—offline after a direct order from the White House. According to Axios, senior company leaders are currently in Washington, D.C. on Sunday to negotiate with the Trump Administration. The export control order, delivered via letter Friday evening, required Anthropic to disable the models or face unspecified consequences. The models are part of the "Mythos-class" lineage, built on the same core technology as the notoriously powerful Mythos Preview, which Anthropic itself deemed too dangerous to release in April.

The takedown followed reports of successful jailbreaks from Amazon and five other companies, who demonstrated to the White House on Thursday and Friday that they could bypass the models' extensive guardrails. An additional Semafor report raised concerns that a China-linked group had gained access to a Mythos-class model, though Anthropic blocks access from China. The White House gave Anthropic just 90 minutes on Friday afternoon to disable the models, citing an unspecified national security threat. Anthropic's response has reportedly been insufficient, with the administration unsatisfied with the company's handling of the jailbreak complaints. The incident marks one of the most aggressive federal interventions into AI deployment, with the Trump Administration taking a hard line on frontier model security.