Anthropic has taken its AI models Claude Fable 5 and Mythos 5 offline following a US government export control directive issued on June 12, 2026, citing national security concerns about a possible method to bypass protections in the models. The company removed access to the models for all users to comply with the order.
What Happened
On June 12, 2026, Anthropic announced it was disabling access to Claude Fable 5 and Mythos 5 after receiving a government letter at 5:21 pm Eastern Time instructing the company to suspend access for all foreign nationals, including foreign national employees. The directive was based on concerns of a discovered technique that could “jailbreak” Claude Fable 5 to bypass its built-in safeguards.
Key Facts
- Company: Anthropic, an artificial intelligence startup known for its Claude AI models
- Models Affected: Claude Fable 5 and Mythos 5
- Directive Received: June 12, 2026, from the US government’s export control office
- National Security Concerns: Method allegedly discovered that could bypass model safeguards (“jailbreaking”)
- Previous Government Action: US Department of Defense designated Anthropic a “supply chain risk” earlier in 2026, restricting military and contractor use of its AI
- Model Release Timeline: Mythos Preview limited rollout in April 2026; Claude Fable 5 publicly launched on June 9, 2026
- Safeguards: Claude Fable 5 was designed to restrict responses on cybersecurity, biology, and chemistry topics
Why It Matters
The government action highlights escalating scrutiny on AI companies over national security risks, particularly how AI models could be exploited to develop cyberattack tools. Anthropic’s compliance affects users relying on these models for cybersecurity advancements and illustrates ongoing tensions between AI innovation and regulatory oversight.
Background
Anthropic has been navigating conflicts with the Trump administration’s national security concerns since early 2026. The US Department of Defense labeled Anthropic a supply chain risk after the company set ethical boundaries restricting military applications of its technology. Anthropic responded with lawsuits challenging these government restrictions.
Claude Fable 5 was developed as a safer version of Mythos, aiming to enable cybersecurity uses while limiting potentially harmful outputs. These models were introduced to balance innovation with preventing misuse in sensitive areas like hacking and biochemistry.
Analysis
Anthropic stated in a blog post that the government’s evidence of the jailbreak was limited to a single demonstration involving minor, previously known vulnerabilities that other public AI models can also identify without bypassing safeguards. The company argued the jailbreak’s impact is narrow and does not significantly increase risks compared to other models.
CEO Dario Amodei previously emphasized supporting a transparent and fair government process for AI safety, but called the current government action inconsistent with these principles.
Who Is Affected
- All Anthropic customers currently using Claude Fable 5 and Mythos 5 models
- Foreign nationals within and outside the United States, including Anthropic’s foreign national employees
- US government agencies and contractors who were already barred from using Anthropic technology under prior designations
- Organizations relying on AI for cybersecurity improvements who had been testing these models
What Remains Unclear
- The specific details and technical nature of the government’s national security concerns were not disclosed in the directive
- Whether other potential jailbreak techniques exist beyond the single demonstration shared with Anthropic
- The duration of the imposed export control and conditions for restoring model access
- The official response from the White House and Commerce Department remains unavailable
What Comes Next
Anthropic will maintain the models offline until further notice. No timeline or conditions for resuming access have been announced publicly. The company’s ongoing legal challenges to government restrictions may also influence future developments.
Sources
This article is based on reporting and publicly available information from the following source:
Read more US News stories on Goka World News.