short
- The US government has ordered Anthropic to suspend access to its most advanced AI models, Fable 5 and Mythos 5, citing national security concerns over a potential jailbreak vulnerability.
- Anthropic complied but disputed the finding, arguing that the vulnerability was simple and already replicable using other publicly available models such as GPT-5.5.
- The company warned that the directive sets a dangerous precedent that, if applied industry-wide, could halt all deployments of new frontier AI models.
The US government issued emergency export control guidance on Friday Anthropy asked Immediately suspends access to the two most powerful AI models – Cloud Fable 5 and Cloud Mythos 5, both Released just a few days ago– To all foreign nationals, including company employees, due to national security concerns.
The directive prohibits any foreign national from accessing the models, whether inside or outside the United States. The extensive scope of the order forced Anthropic to disable forms for its entire customer base to ensure compliance.
The letter did not provide specific details about its national security concerns, but the government believes it has become aware of a method to bypass or “jailbreak” the publicly available Fable 5 model. Mythos 5, which has fewer guardrails and is particularly powerful at detecting cybersecurity vulnerabilities, was only available to select partners.
Anthropist doubted the seriousness of this outcome. The company said it reviewed a demonstration of this technology and concluded that the vulnerabilities identified appear to be relatively minor, and that other publicly available models are also able to detect them without any overriding required.
The government has so far only provided verbal evidence of the potential for a limited, non-universal jailbreak, which essentially consists of requiring the model to read a specific code base and fix any software flaws, the company said. Anthropic added that it has demonstrated that the level of capability on offer is already available on a wide range of competing models, including OpenAI’s GPT-5.5.
While adhering to the directive, Anthropic said it believed the action set a dangerous precedent. “If this standard were implemented across the industry, we believe it would essentially halt all new model deployments for all frontier model providers,” the company wrote.
Access to all other humanitarian models will not be affected. The company said it is working to restore access as soon as possible.
I’ve had a number of conversations with people inside and outside of government about the current situation with Anthropics, and this is what I believe to be true:
-As we know, Anthropic publicly released their Mythos-class models earlier this week under the brand name Fable….
– David Sacks (@DavidSacks) June 13, 2026
On Saturday, David Sachs — co-chair of the President’s Council of Advisors on Science and Technology —Written on X “A highly credible and trusted partner of both Anthropic and the US government, who was testing Fable, offered to jailbreak those security barriers. The moderator (Anthropic CEO, Dario Amodei) requested that the jailbreak be fixed or the model be undeployed. Dario declined.”
Sachs further claimed that Anthropic’s response to the government’s request conflicted with the company’s public claims regarding the need for safety and regulation of AI — a perspective that Amodei again reiterated. Subscribed this week In a blog post.
“Anthropic has prioritized the continued display of the consumer model over safety,” Sacks wrote. “In response, the administrator issued export controls. The administrator did so reluctantly. We were very surprised that Anthropic did not want to cooperate with a reasonable security request (i.e. fix the jailbreak issue). Anthropic’s reaction is completely inconsistent with its brand and ethos as a secure AI research community.”
Sachs added that management hopes Anthropic will fix the issue and that Fable 5 will be able to return to general release.
“The supervisor appreciates Anthropic’s technical capabilities and feels that this issue, although serious, should be easily resolved,” he wrote. “The ball is in the Anthropists’ court.”
Earlier this year, Anthropic and the US government clashed when the AI company refused to sign an expansive agreement allowing mass domestic surveillance of Americans and fully autonomous lethal weapons systems.
President Donald Trump Criticize the company After the talks collapsed, the Ministry of Defense described Anthropic as a “supply chain risk” – a designation made by the company Appeal in court. Since then, there have been reports of a confrontation between the two sides I got cold The government sought to use Cloud Mythos and other models.
Daily debriefing Newsletter
Start each day with the latest news, plus original features, podcasts, videos and more.
