The Trump administration ordered Anthropic to remove its Mythos and Fable AI models within 90 minutes on Friday, June 12, 2026, citing an unspecified national security threat. The company complied, beginning the de-deployment of the models before receiving formal Commerce Department controls later that afternoon, according to Yahoo News.
Government Intervention and Model Removal
The directive originated from a phone call at 1:00 p.m. ET on Friday. According to an Anthropic source, the government demanded the immediate withdrawal of the Mythos and Fable models. While the government characterized the models as a “national security threat,” it provided no further details to the company during the exchange.
Anthropic had previously engaged in pre-release testing of the Fable model with government officials and had secured explicit approval for its deployment prior to the demand. Despite this, the company reported that the government held firm on its 90-minute deadline. By 5:30 p.m. ET that same day, Anthropic received a letter from the Commerce Department establishing a new licensing regime and imposing strict controls on the usage and accessibility of the models. This action aligns with the broader authority of the U.S. Department of Commerce’s Bureau of Industry and Security (BIS), which manages export controls and technology restrictions based on the Export Administration Regulations (EAR).
Regulatory Context and Security Concerns
The sudden demand highlights the volatile regulatory environment surrounding high-capacity AI systems. Reporting indicates that the move was part of a broader, rapid government effort to address warnings regarding model capabilities from various stakeholders, including Amazon. The intervention mirrors the growing government focus on “frontier models,” which are defined by their advanced capabilities that could potentially be misused for biological, chemical, or cyber-offensive operations.
This incident follows a period of heightened scrutiny regarding Anthropic’s safety protocols. In November 2025, the company reported that hackers sponsored by the Chinese government had bypassed safety safeguards to use Claude for automated cyberattacks against approximately 30 global organizations. Following that event, Anthropic announced in September 2025 that it would cease selling products to groups majority-owned by entities in China, Russia, Iran, or North Korea. These restrictions were intended to serve as a voluntary baseline for the industry, but the June 2026 mandate suggests a shift toward compulsory federal oversight rather than industry-led self-regulation.
The use of emergency regulatory powers, such as the Defense Production Act or specific Commerce Department directives, is historically reserved for scenarios where the government determines that a product poses an “imminent threat” to national security. In the context of artificial intelligence, regulators have increasingly looked toward the “compute threshold” as a trigger for oversight, monitoring the amount of processing power used to train and run large-scale models.
Anthropic’s Current Operational Status
As of June 2026, Anthropic remains the world’s most valuable pure-play AI company, with an estimated valuation of $965 billion. Founded in 2021 by former OpenAI employees, including CEO Dario Amodei and President Daniela Amodei, the firm has focused heavily on AI safety and the development of the Claude series of large language models. The company’s architecture is built on the principle of “Constitutional AI,” a framework designed to align model behavior with a set of written principles to reduce harmful outputs.
The company’s growth has been supported by significant infrastructure investments. In October 2025, Anthropic secured a cloud partnership with Google for access to one million custom Tensor Processing Units. Additionally, Microsoft and Nvidia were expected to invest up to $15 billion in the company in late 2025, alongside a $30 billion compute capacity agreement with Microsoft Azure. These partnerships are critical to the company’s operations, as the high-intensity computing requirements of models like Mythos and Fable necessitate massive, specialized server infrastructure.
The mandate to pull the Fable and Mythos models marks a significant shift in the relationship between the company and federal regulators. When asked about the nature of the threat, the company maintained its commitment to safety protocols.
We immediately sought to understand the specific nature of the threat so we could remediate it.
The imposition of a new licensing regime by the Commerce Department suggests that Anthropic—and potentially other AI developers—will face a more rigorous, pre-deployment review process moving forward. Such regimes typically require companies to submit “red-teaming” reports to federal agencies for evaluation before a model is made publicly available. The speed of the June 12 action indicates that federal agencies are prioritizing real-time intervention over traditional, slower administrative notice-and-comment procedures, placing the burden of compliance squarely on the developers of large-scale AI systems.
Find more reporting in our Business section.
