The U.S. government forced Anthropic to take its Claude Mythos 5 AI model offline this week due to catastrophic security risks [1].

This intervention marks a significant escalation in federal oversight of artificial intelligence. By disabling one of the most powerful models available, the administration is prioritizing national security and the prevention of systemic failures over the commercial availability of frontier AI.

The shutdown occurred after Anthropic released a safeguarded version of the technology called Fable 5 [1]. Despite the existence of this safer alternative, officials said the original Mythos 5 model posed severe jailbreak risks that could lead to catastrophic outcomes [2].

Claude Mythos 5 is a 10 trillion-parameter model [3]. The sheer scale of the system contributed to the government's concerns regarding its stability and the potential for malicious actors to bypass its internal safety protocols.

Reports on the mechanism of the shutdown vary slightly. Some sources said the U.S. government forced the model offline [4], while others indicate that export-control actions taken by the Trump administration effectively shut the system down [5].

Anthropic has not provided a timeline for when, or if, the full Mythos 5 model will return to public access. The company continues to offer Fable 5 as the primary alternative for users seeking high-capacity AI capabilities [1].

“The U.S. government forced Anthropic to take its Claude Mythos 5 AI model offline”

This move signals a shift toward more aggressive government intervention in AI development. By using export controls and direct mandates to disable a 10 trillion-parameter model, the U.S. is establishing a precedent where national security concerns can override a company's right to deploy its own software, regardless of whether a 'safeguarded' version exists.