Anthropic Suspends Powerful AI Model Amid US Security Order
In a dramatic move, Anthropic has halted the operation of its newly released Claude Fable 5, following a security order from U.S. authorities. The order demands that the company cease allowing foreign nationals to access the model, citing concerns over “jailbreaking” techniques that could bypass safeguards.
According to a statement on its website, Anthropic was told “we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance.” The firm added that security officials had not identified specific problems, but believed the government had detected a method to exploit the system.
Fable 5, a successor to the company’s Claude Mythos platform, positioned itself as rival to OpenAI’s ChatGPT and Google’s Gemini. The model was praised for its “intelligence” but raised warnings from finance, technology and government leaders. In April, Anthropic granted pre‑release access to a slim group of organisations to test vulnerabilities, describing the model as “too powerful to release” in its public debut.
The pause comes amid a feud with the Trump administration. Former U.S. Secretary of Defense Pete Hegseth labeled Anthropic a “supply chain risk” – a designation normally reserved for firms perceived as threats in adversarial nations. The Pentagon had issued a directive to prevent government agencies from using the company's tools, but a federal judge ruled the order could not be enforced, allowing continued access for agencies still required to work with Anthropic.
Anthropic’s case underscores a growing debate over AI governance. While the firm touts safeguards against hacking, officials point to potential “jailbreaking” that could expose sensitive data. The broader implication is a cautionary spotlight on how AI that drives climate models, environmental forecasting and sustainability strategies must be balanced against national security safeguards.
