Core Summary
Anthropic has suspended all customer access to its latest AI models, Claude Fable 5 and Mythos 5, after US national security authorities identified “jailbreak” vulnerabilities. The company stated it was “ordered to suspend foreign nationals from using Claude Fable 5” and that “the net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance.” This marks the latest escalation in Anthropic’s ongoing conflict with the Trump administration, after the Pentagon labeled the company a “supply chain risk.”
Event Details
In a statement published on its website, Anthropic revealed that the US government raised national security concerns about a method of bypassing or “jailbreaking” Fable 5’s safety restrictions. Jailbreaking refers to the process of getting past software restrictions designed to protect cyber networks, allowing hackers to access sensitive information or unblock features.
“Our understanding is that the government believes it has become aware of a method of bypassing, or ‘jailbreaking’ Fable 5,” Anthropic said. “We reviewed a demonstration of this specific technique being used to identify a small number of previously known, minor vulnerabilities. These vulnerabilities all appear relatively simple, and we have found that other publicly-available models are able to discover them as well without requiring a bypass.”
Panoramic Analysis
This incident marks an unprecedented escalation in AI safety regulation. Before releasing Fable 5, Anthropic proudly declared the model “too powerful to release” and implemented multiple safeguards. However, the UK government’s AI Security Institute found that the model could exploit defenses and systems 73% of the time, indicating significant vulnerabilities even in top AI companies’ safety measures.
The deeper issue is that this is not merely a technical security matter but a political confrontation. Anthropic’s conflict with the Trump administration has escalated to legal action: the Pentagon publicly labeled Anthropic a “supply chain risk” for the first time ever for a US company — a designation historically reserved for companies based in adversarial countries. Anthropic is suing the Pentagon, and a US judge has ruled that the Pentagon’s directive cannot be enforced while the lawsuit continues.
The European Union responded swiftly, emphasizing that this development “further underlined Europe’s need for technological sovereignty.” The European Commission this month unveiled measures to slash the 27-nation bloc’s dependence on America and Asia for key technologies, including AI.
Gina Neff, Professor of Responsible AI at Queen Mary University London, warned that restricting model access could limit the development and safe testing of AI systems and restrict collaboration with governments worldwide. “We’re in uncharted territory at this point,” she said. “People within the AI industry have been warning us that these tools are getting better very rapidly and that we have to be able to build up capabilities to keep our companies safe from cyber attacks.”
Multiple Perspectives
Anthropic’s Position: The company emphasized its safeguards, claiming the discovered vulnerabilities are “relatively simple” and discoverable by other models, suggesting the government’s concerns may be overstated.
US Government’s Position: Imposing strict controls on national security grounds, treating the most advanced AI models as strategic assets equivalent to military technology.
EU Response: Viewing this incident as evidence for reducing dependence on American technology, accelerating its technological sovereignty strategy.
Academic Perspective: Professor Gina Neff expressed concern that restrictions could be counterproductive, hindering global AI safety research and collaboration.
Editor: GoodInfo Global News Team