US Government Orders Halt to Anthropic’s Fable 5, Mythos 5 AI Models Over Security Concerns

WASHINGTON – The U.S. government has issued a directive ordering artificial intelligence company Anthropic to suspend all access to its Fable 5 and Mythos 5 models, citing unspecified national security authorities. The order impacts foreign nationals and foreign national employees of Anthropic, effectively requiring the company to disable the models for all customers to ensure compliance.

Anthropic received the directive on June 12, 2026, at 5:21 p.m. ET. The company stated that the government believes a method for bypassing or “jailbreaking” Fable 5 has been discovered. However, Anthropic asserts that the identified vulnerabilities are minor and present in other publicly available AI models, questioning the severity of the government’s action.

AI Model Vulnerabilities and Government Response

The directive stems from the government’s belief that a specific technique can bypass Fable 5’s security protocols. Anthropic, in a statement released on its website, acknowledged reviewing a demonstration of this technique. The company maintains that the identified vulnerabilities are relatively simple and that other models, including publicly available ones, can also identify them without requiring a bypass mechanism. Anthropic’s internal testing indicated that while non-universal jailbreaks are possible, a universal jailbreak that broadly bypasses safeguards has not been found.

Anthropic has implemented safeguards designed to mitigate misuse, particularly in cybersecurity-related tasks. The company stated that these safeguards are considered more effective than those of previously deployed models. Despite this, the government’s directive indicates a concern that these safeguards may not be sufficient against evolving exploitation methods. The exact nature of the national security concern was not detailed in the government’s written communication to Anthropic.

Anthropic’s Stance and Industry Implications

Anthropic expressed disagreement with the government’s decision, arguing that recalling a commercially deployed model based on a narrow, non-universal jailbreak finding sets a precedent that could halt the deployment of new AI models across the industry. The company advocates for a transparent, fair, and technically grounded statutory process for blocking unsafe AI deployments, suggesting the current action does not adhere to these principles.

The company is complying with the legal directive but aims to restore access to Fable 5 and Mythos 5 as soon as possible, believing the situation is a misunderstanding. Access to other Anthropic models, such as Claude, will not be affected by this order. The directive mandates the retention of customer data for 30 days for Fable 5, a policy Anthropic implemented to research and mitigate jailbreaks, incurring significant costs.

Global AI Policy and Regulatory Scrutiny

This action by the U.S. government places increased scrutiny on the development and deployment of advanced AI models. Regulators worldwide are grappling with the potential national security risks posed by powerful AI technologies, including concerns about sophisticated cyberattacks and autonomous weapon systems.

The U.S. directive signals a more assertive stance in controlling access to frontier AI capabilities, particularly when perceived vulnerabilities could be exploited by state or non-state actors. Other nations and international bodies are closely monitoring such developments, which could influence future global AI governance frameworks and trade policies related to sensitive technologies.

The differing perspectives between Anthropic and the U.S. government highlight the ongoing debate about the appropriate balance between fostering AI innovation and ensuring robust security. While AI developers emphasize the inherent limitations of current safeguards and the constant cat-and-mouse game of cybersecurity, governments are prioritizing the prevention of potential large-scale harms.

The eventual resolution of this dispute and Anthropic’s efforts to restore access will likely set important benchmarks for regulatory engagement with the AI industry, particularly concerning national security implications. The company has committed to providing further details within 24 hours.