Anthropic Disables AI Models After US Government Order

Anthropic suspended access to its Claude Mythos 5 and Claude Fable 5 AI models on June 26, 2024 [3, 6].

The move signals a tightening of U.S. government oversight regarding how advanced artificial intelligence is distributed globally. By restricting these specific models, the U.S. is attempting to prevent high-capability AI from falling into the hands of foreign adversaries who might repurpose the technology for malicious use.

The suspension followed an export-control order issued by the U.S. government [1, 4]. Officials said national security concerns were the primary driver for the directive [1, 5]. Specifically, the government said the models could be "jailbroken" to bypass the safety guardrails installed by Anthropic [1, 5].

Export controls are designed to limit the transfer of sensitive technologies to certain foreign nationals, or entities ^[3]. In this case, the U.S. government determined that the capabilities of the Mythos 5 and Fable 5 models posed a risk if accessed without strict oversight [1, 2].

European regulators have begun evaluating the implications of these curbs ^[2]. Because AI development is a global effort, a U.S. directive that limits access to specific models can disrupt the operations of developers and companies across other continents. This creates a tension between national security priorities in Washington and the open-market goals of the global tech industry.

Anthropic has not provided a timeline for when, or if, these models will be restored for general use. The company acted to comply with the government order ^[5]. This incident highlights the increasing role of state intervention in the deployment of large language models as they move from experimental tools to critical infrastructure.