Anthropic’s Fable Is Locked Down As US Takes AI Safety Into Its Hands

The White House has ordered the immediate shutdown of Anthropic's advanced AI system, codenamed 'Fable,' marking the first time the U.S. government has forcibly halted a major AI deployment on national security grounds. The directive, issued under the Defense Production Act, restricts access to the model exclusively to U.S. nationals, citing risks of foreign exploitation and algorithmic sabotage. The move pits Anthropic's founding mission of responsible AI development against a new era of government-mandated controls, raising urgent questions about who decides what safe AI looks like.

Anthropic, co-founded by former OpenAI researchers Dario Amodei and Daniela Amodei, has long positioned itself as the safety-first alternative in the AI arms race. Its flagship model Claude is famous for its constitutional AI approach. Fable was reportedly a next-generation reasoning engine designed for critical infrastructure applications. The White House's action, detailed in a classified directive first revealed by Forbes, stunned the industry because it targeted a company known for its safety advocacy rather than a less scrupulous player.

The context is a rapidly shifting regulatory landscape. The Biden administration's 2023 Executive Order on AI mandated safety testing for frontier models. In 2025, Congress passed the AI Accountability Act, creating a new oversight body. But this shutdown goes further—it is an operational takedown, not a testing requirement. The directive specifically prohibits any non-U.S. person from accessing Fable's weights, training data, or API endpoints, effectively nationalizing control of the technology.

According to sources familiar with the order, the White House acted after intelligence agencies flagged that Fable's underlying architecture could be reverse-engineered by adversarial states to create 'undetectable AI weapons.' The directive was signed by the National Security Advisor and implemented by the Department of Commerce's Bureau of Industry and Security. Anthropic was given 72 hours to comply, and its servers were sealed by federal agents on June 13, 2026.

The implications are profound. First, it creates a dangerous precedent: a U.S. company's sovereign control over its own AI can be overridden by executive fiat. Second, it fractures the global AI market—Fable was designed for deployment in allied nations like the UK and Japan. Third, it challenges Anthropic's core identity. CEO Dario Amodei released a statement saying, 'We built Fable to demonstrate that safety and capability can coexist. This decision suggests the government doubts that very premise.' Critics argue the move is an overreach that stifles innovation and punishes the most responsible actors.

What happens next will reverberate for years. Anthropic is likely to challenge the directive in federal court, arguing it violates the Fifth Amendment's due process clause. Congress has scheduled hearings for July, and the AI Safety Board will release a public report within 90 days. Meanwhile, other frontier labs—OpenAI, Google DeepMind, and xAI—are watching closely, as their own models could face similar restrictions. The Fable shutdown may be the moment AI safety advocacy collides irrevocably with state power, redrawing the line between corporate responsibility and national security.