- Anthropic has suspended Claude Fable 5 following an order affecting foreign access to the model.
- US authorities reportedly raised concerns over a potential method for bypassing the system's safeguards.
- The move comes as Anthropic continues its legal battle with the Pentagon over government use of its AI tools.
Anthropic has suspended access to its latest artificial intelligence model, Claude Fable 5, after intervention from US authorities raised fresh concerns about the security risks posed by increasingly powerful AI systems.
The decision places Anthropic AI and AI security concerns firmly back in the spotlight, just days after the company rolled out the model to the public. The move also adds another layer of tension to an ongoing dispute between the AI firm and the Trump administration over the use of its technology within government agencies.
From breakthrough launch to sudden shutdown
In a statement published on its website, Anthropic said it had been instructed to restrict access to Claude Fable 5 for foreign nationals. According to the company, complying with the order required it to disable both Fable 5 and Mythos 5 across its customer base.
The company said national security officials had not provided detailed explanations for the action. However, Anthropic said its understanding was that authorities had become aware of a possible technique for bypassing, or "jailbreaking", the model's safeguards.
Jailbreaking refers to attempts to circumvent restrictions built into software systems, potentially allowing users to uncover vulnerabilities, access protected information or unlock functions that were not intended to be available.
Anthropic said it had reviewed demonstrations of the technique and found only a small number of previously known weaknesses. The company reportedly said these vulnerabilities were relatively minor and that similar publicly available AI models were capable of identifying them without requiring any form of jailbreak.
Security fears collide with AI ambitions
The suspension comes after months of debate around the risks associated with advanced AI systems.
Before launching Claude Fable 5, Anthropic repeatedly highlighted the safeguards built into the model and acknowledged concerns surrounding its ability to identify vulnerabilities in computer systems. The company granted a limited number of organisations early access in April to test the technology and assess potential risks before a wider release.
Anthropic had also attracted attention by describing the model as exceptionally powerful. According to the company, as quoted in a news report, Fable 5's capabilities surpassed those of any model it had previously made available to the public. Critics, however, questioned whether such characterisations reflected genuine concern or aggressive marketing.
The latest development arrives against the backdrop of an increasingly strained relationship between Anthropic and the Trump administration.
President Donald Trump has publicly criticised the company, while former Defence Secretary Pete Hegseth designated Anthropic a "supply chain risk", reportedly making it the first US technology company to receive such a classification. The designation is typically associated with concerns over whether a product or service can be trusted for government use and has historically been applied to entities linked to rival nations.
Anthropic is currently challenging that designation in court. A federal judge has ruled that the Pentagon's directive cannot be enforced while the case proceeds, allowing government departments and military contractors to continue using the company's AI tools for the time being.
The abrupt withdrawal of Claude Fable 5 is likely to fuel wider debate over how governments should regulate increasingly capable AI systems. While officials have yet to publicly outline the precise concerns behind the intervention, the episode highlights the growing collision between technological advancement, national security and the race to build more powerful AI models.
