Blab AI
Anthropic Disables Fable 5 and Mythos 5 Models Amid Trump Export Control Directive
In a swift move prompted by a late‑Friday letter from the U.S. government, Anthropic, the San Francisco‑based AI firm, shut down two of its most advanced language models—Fable 5 and Mythos 5—for all customers on June 12. The company’s decision follows a directive from the Trump administration that requires the models to be suspended for any foreign national, regardless of whether the user is inside or outside the United States. Access to Anthropic’s other models, including the Claude series, remains unchanged.
The export‑control letter did not spell out the national‑security concerns that underlie the order. According to Anthropic, the government believes there is a way to bypass, or “jailbreak,” Fable 5. The firm described the potential jailbreak as a narrow, non‑universal technique that would involve asking the model to read a specific codebase and correct software flaws. Anthropic said that a potential jailbreak was shared with the government, but that the evidence was verbal and not formally documented.
The directive follows a long‑standing dispute between Anthropic and the Department of Defense (DoD) over the company’s contractual restrictions on military use of its models. Anthropic has refused to remove clauses that prohibit mass domestic surveillance or fully autonomous weapons. In response, DoD Secretary Pete Hegseth designated the company a “supply‑chain risk,” a label typically reserved for foreign national security threats. Anthropic has filed two lawsuits alleging that the government’s actions constitute illegal retaliation.
The move signals a clear shift in the Trump administration’s approach to AI export controls. In May 2025, the president rescinded the Biden‑era “AI diffusion rule,” which had limited the export of advanced AI chips. The new directive marks a departure from the administration’s prior hands‑off stance on AI model exports.
While Anthropic is complying with the legal order, the company disputes the government’s reasoning. It argues that a narrow potential jailbreak should not justify recalling a commercial model that has already been deployed to hundreds of millions of users. Anthropic warned that applying the same standard industry‑wide would effectively halt new model deployments for all frontier model providers. The firm called for a statutory process that is transparent, fair, clear, and grounded in technical facts, and said the current action does not meet those principles.
The decision has drawn criticism from technology experts and nonprofit groups. Americans for Responsible Innovation (ARI), whose president is Brad Carson, said the regulation should be applied consistently across the industry, without favoritism, and based on a clear rules‑based process. Carson warned that the decision to block Anthropic’s models fails that test and could risk the United States’ edge in AI innovation.
As of now, Anthropic’s other models, including the Claude series, remain available to customers. The company’s lawsuits against the DoD and the Trump administration are ongoing, and the legal outcome could influence future export‑control policies. Analysts note that the incident highlights the growing tension between AI developers and government regulators over national‑security concerns and the potential for model misuse.
The broader AI industry watches closely. The directive may prompt other companies to review their export‑control compliance and contractual clauses. It also raises questions about how export controls will be applied to future models, especially those with capabilities that could be leveraged for software vulnerability detection or other sensitive applications.
In summary, Anthropic has temporarily disabled Fable 5 and Mythos 5 in response to a Trump‑era export‑control directive that cites a potential jailbreak as the basis for the restriction. The company is complying with the order but disputes its justification, and the move has sparked legal challenges and industry debate over the scope of AI export controls and the balance between innovation and national security.
The export‑control letter did not spell out the national‑security concerns that underlie the order. According to Anthropic, the government believes there is a way to bypass, or “jailbreak,” Fable 5. The firm described the potential jailbreak as a narrow, non‑universal technique that would involve asking the model to read a specific codebase and correct software flaws. Anthropic said that a potential jailbreak was shared with the government, but that the evidence was verbal and not formally documented.
The directive follows a long‑standing dispute between Anthropic and the Department of Defense (DoD) over the company’s contractual restrictions on military use of its models. Anthropic has refused to remove clauses that prohibit mass domestic surveillance or fully autonomous weapons. In response, DoD Secretary Pete Hegseth designated the company a “supply‑chain risk,” a label typically reserved for foreign national security threats. Anthropic has filed two lawsuits alleging that the government’s actions constitute illegal retaliation.
The move signals a clear shift in the Trump administration’s approach to AI export controls. In May 2025, the president rescinded the Biden‑era “AI diffusion rule,” which had limited the export of advanced AI chips. The new directive marks a departure from the administration’s prior hands‑off stance on AI model exports.
While Anthropic is complying with the legal order, the company disputes the government’s reasoning. It argues that a narrow potential jailbreak should not justify recalling a commercial model that has already been deployed to hundreds of millions of users. Anthropic warned that applying the same standard industry‑wide would effectively halt new model deployments for all frontier model providers. The firm called for a statutory process that is transparent, fair, clear, and grounded in technical facts, and said the current action does not meet those principles.
The decision has drawn criticism from technology experts and nonprofit groups. Americans for Responsible Innovation (ARI), whose president is Brad Carson, said the regulation should be applied consistently across the industry, without favoritism, and based on a clear rules‑based process. Carson warned that the decision to block Anthropic’s models fails that test and could risk the United States’ edge in AI innovation.
As of now, Anthropic’s other models, including the Claude series, remain available to customers. The company’s lawsuits against the DoD and the Trump administration are ongoing, and the legal outcome could influence future export‑control policies. Analysts note that the incident highlights the growing tension between AI developers and government regulators over national‑security concerns and the potential for model misuse.
The broader AI industry watches closely. The directive may prompt other companies to review their export‑control compliance and contractual clauses. It also raises questions about how export controls will be applied to future models, especially those with capabilities that could be leveraged for software vulnerability detection or other sensitive applications.
In summary, Anthropic has temporarily disabled Fable 5 and Mythos 5 in response to a Trump‑era export‑control directive that cites a potential jailbreak as the basis for the restriction. The company is complying with the order but disputes its justification, and the move has sparked legal challenges and industry debate over the scope of AI export controls and the balance between innovation and national security.
