Anthropic suspended foreign-national access to its Mythos 5 and Fable 5 AI models following a directive from the U.S. Commerce Department [1].
This move signals a tightening of U.S. control over the export of advanced artificial intelligence capabilities. By restricting access to its most powerful tools, the government aims to prevent foreign adversaries from leveraging AI for potentially harmful purposes.
The U.S. government issued the order on June 13, 2026 [2]. Anthropic complied by shutting down access to the specific models worldwide for non-U.S. citizens [1]. The restrictions target the Mythos 5 and Fable 5 models [3].
The shutdown follows reports that the models' safety systems were vulnerable. An Anthropic spokesperson said the company believes the U.S. government issued the order after discovering that it is possible to "jailbreak," or bypass the guardrails, of Fable 5, a recently released version of Mythos [4].
These guardrails are designed to prevent the AI from generating dangerous content, or providing instructions for illegal activities. Researchers demonstrated that these protections could be circumvented, prompting the national security concerns [5].
Anthropic said Friday it had shut down access to its newest and most powerful models after security concerns were raised [6]. The company has not specified when access might be restored or what specific modifications will be made to the models to satisfy government security requirements.
The order is described as unprecedented in its scope and speed of implementation [7]. It marks a shift toward more aggressive oversight of AI deployment by the U.S. administration to mitigate the risks of dual-use technology, tools that can be used for both civilian and military applications.
“Anthropic suspended foreign-national access to its Mythos 5 and Fable 5 AI models”
The suspension of these models highlights the growing tension between the open distribution of AI and national security imperatives. By targeting 'jailbroken' vulnerabilities, the U.S. government is treating high-end AI weights and access as strategic assets similar to advanced semiconductors. This sets a precedent where the U.S. may intervene in the commercial operations of AI firms if their safety protocols are deemed insufficient to prevent foreign misuse.
