Anthropic, a prominent artificial intelligence developer, has disabled some access to its Fable 5 and Mythos 5 AI models. The action, taken on Friday, June 12, was a direct response to a U.S. government export control directive. This directive, as detailed in Anthropic’s announcement, cited unspecified “national security authorities” as the basis for its mandate.
The government’s order specifically instructed Anthropic to suspend access to these advanced models by “any foreign national,” a broad category that includes individuals both within and outside the U.S., as well as Anthropic’s own employees. The core concern articulated by the government, according to Anthropic, is a belief that a method for bypassing, or ‘jailbreaking,’ the Fable 5 model has been discovered.
Anthropic acknowledged reviewing a demonstration of this specific technique, noting it was used to identify a small number of previously known, minor vulnerabilities. The company stated that these vulnerabilities appeared relatively simple and that other publicly available models could discover them without requiring such a bypass. Despite complying with the government’s directive, Anthropic expressed its disagreement with the premise that a “narrow potential jailbreak” justifies recalling a model that had been deployed to hundreds of millions of users.
The AI firm further contended that if this standard were applied universally across the industry, it would effectively halt all new model deployments for all frontier model providers. This stance highlights a significant tension between national security imperatives and the rapid development and deployment cycles characteristic of the AI sector.
Adding another layer to the government’s intervention, a report by The Wall Street Journal indicated that the directive was prompted by discussions between Amazon CEO Andy Jassy and U.S. officials. An Amazon spokesman, commenting on the matter to the WSJ, stated, “As a leading cloud provider that serves a large number of private and public sector customers, it’s not uncommon for governments to seek our counsel on potential security risks.” The spokesman added that Amazon does not share the details of such discussions when they occur.
This incident is not Anthropic’s first encounter with governmental scrutiny. Earlier this year, the Defense Department declared the company a supply chain risk, a designation that bars companies working with the military from engaging with the AI startup. This followed previous pushback from Anthropic regarding the military’s potential use of its technology for mass domestic surveillance and autonomous weapons systems. Anthropic has since initiated legal proceedings against the government to overturn this supply chain risk designation.
More broadly, Anthropic has been an advocate for robust governmental oversight of AI. The startup has called on world governments to address the inherent risks posed by AI, specifically by securing the authority to prohibit dangerous deployments. Furthermore, Anthropic has urged governments to prepare workers and the broader economy for the transformative impact of AI and to ensure that the financial benefits derived from the technology are equitably distributed.
The suspension of access to Fable 5 and Mythos 5 underscores the escalating complexities at the intersection of advanced AI development, national security concerns, and the evolving regulatory landscape, signaling a critical period for the future of frontier AI models.
