Anthropic Suspends New AI Models Following US Security Concerns
Artificial intelligence company Anthropic has suspended access to its recently launched AI models, Claude Fable 5 and Mythos 5, after receiving directives from US authorities over potential national security concerns.
The company announced the decision in a statement published on its website, revealing that restrictions imposed by the US government have forced it to disable access to the tools for customers worldwide.
According to Anthropic, the move follows concerns raised by government officials regarding the possibility that the newly released models could be vulnerable to advanced jailbreak techniques capable of bypassing built-in safeguards.
“The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance,” the company stated.
Anthropic noted that it had not been informed of any specific security breach but said authorities believed a method had been discovered that could circumvent some of the models’ protective measures.
“Our understanding is that the government believes it has become aware of a method of bypassing, or ‘jailbreaking’ Fable 5,” the company explained.
The firm said it reviewed demonstrations of the alleged technique and concluded that the vulnerabilities identified were already known and relatively minor. It also maintained that similar weaknesses could be found using other publicly available AI models without requiring any specialised bypass methods.
Prior to the public release of Claude Fable 5, Anthropic had repeatedly highlighted the model’s advanced capabilities and the extensive safeguards built into the system. The company had initially limited access to a small number of organisations during a testing phase to assess potential risks and identify vulnerabilities before launch.
Anthropic previously described the technology as exceptionally powerful, arguing that its advanced capabilities required careful oversight due to its potential ability to discover and exploit weaknesses in computer systems.
The suspension comes amid growing global debate over the rapid advancement of AI technologies and the need for stronger regulatory frameworks to govern their deployment.
The European Commission responded to the development by saying it reinforced the importance of technological independence within Europe. Officials indicated they were monitoring the situation closely as they continue efforts to reduce reliance on foreign technologies in strategic sectors, including artificial intelligence.
Experts have also expressed concerns about the implications of restricting access to advanced AI systems. Gina Neff, Professor of Responsible AI at Queen Mary University London, warned that limiting access could affect both safety research and international collaboration on AI governance.
She noted that recent testing suggested the models demonstrated significantly improved capabilities in identifying and exploiting cybersecurity weaknesses, describing the development as a major leap forward in AI-driven cyber operations.
The suspension adds to ongoing tensions between Anthropic and the US government. The company is currently engaged in a legal dispute with the Pentagon after being designated a “supply chain risk,” a classification that effectively restricts the use of certain technologies by government agencies.
Anthropic has challenged the designation in court, and a judge has ruled that the restriction cannot be enforced while legal proceedings continue, allowing government entities to continue using the company’s AI products pending the outcome of the case.
The latest development highlights the increasing scrutiny facing AI companies as governments seek to balance innovation with concerns over national security, cybersecurity and public safety.
