Anthropic suspends new AI tools over US government security concerns

Harry Sekulichand

Shiona McCallum,Senior Tech Reporter

Anthropic has suspended its powerful new AI model after US authorities raised security concerns just days following its public release.

"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 wrote.

Anthropic and the Trump administration are involved in a separate ongoing lawsuit over an order to stop government agencies using the company's AI tools. The BBC has approached the US Department of Commerce for comment.

Claude Fable 5 is a version of Anthropic's Claude Mythos, an AI program rivalling competitors OpenAI's ChatGPT and Google's Gemini.

Anthropic said US national security authorities had not identified specific concerns.

"Our understanding is that the government believes it has become aware of a method of bypassing, or 'jailbreaking' Fable 5," the company said.

Jailbreaking is a process of getting past software restrictions designed to protect a cyber network, allowing hackers to access sensitive information or unblock features.

"We reviewed a demonstration of this specific technique being used to identify a small number of previously known, minor vulnerabilities," Anthropic said.

"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."

Ahead of the release of Claude Fable 5, the company touted various "safeguards" it had implemented to prevent cyber hacking.

Finance, technology and government leaders had expressed concerns about its public rollout, following a private release in April for previewing and testing vulnerabilities within its own system.

Anthropic said it enabled pre-release access for a handful of organisations because the tool was so intelligent that it could be dangerous because of its ability to exploit or hack computer systems.

The company self-proclaimed that it was "too powerful to release" before Claude Fable 5 was publicly released, which some critics questioned as inflated hype and marketing spin.

"Fable's capabilities exceed those of any model we've ever made generally available," the company said.

The European Union, which gained access to Mythos earlier in June after weeks of talks, said the latest development further underlined "Europe's need for technological sovereignty".

"We take note of Anthropic's statement and are assessing," said Thomas Regnier, a spokesman for the European Commission, which 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, told the BBC that the decision to restrict access to the model could limit the development and safe testing of these AI systems.

She warned it could also restrict collaboration with governments around the world.

"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

She said the UK government's AI Security Institute found in its tests that the model could exploit defences and systems 73% of the time.

"It's a step change in capability in cyber security," she added.

Anthropic has found itself in the crosshairs of the Trump administration recently.

Donald Trump has criticised the company publicly and then US Defence Secretary Pete Hegseth labelled it a "supply chain risk" – the first time a US company has ever publicly received such a designation.

The designation means a tool or service is not secure enough for government use, and is historically reserved for companies based in adversarial countries.

Anthropic is suing the Pentagon following the designation. A US judge has ruled the Pentagon's directive could not be enforced, meaning government agencies and organisations working with the US military can still use Anthropic while the lawsuit continues.