Anthropic has suspended its powerful new Artificial Intelligence model after US authorities raised security concerns just days following its public release. In a statement, Anthropic said it was ordered to suspend foreign nationals from using Claude Fable 5, a program the company had described as “too powerful”. “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. The US Department of Commerce was approached for comment.
Claude Fable 5 is a version of Anthropic’s Claude Mythos, an Artificial Intelligence program rivalling competitors OpenAI’s ChatGPT and Google’s Gemini. Anthropic said US national security authorities had not identified specific concerns, but its understanding was that the government believed it had become aware of a method of bypassing, or “jailbreaking”, Fable 5. The company said it reviewed a demonstration of the technique being used to identify a small number of previously known, minor vulnerabilities. It added that the vulnerabilities appeared relatively simple and that other publicly available models could discover them without requiring a bypass.
Ahead of the release, Anthropic promoted safeguards designed to prevent cyber hacking, while finance, technology and government leaders had expressed concerns about a public rollout. The company had enabled pre-release access for a handful of organisations because the tool was considered intelligent enough to be dangerous through its ability to exploit or hack computer systems. The European Commission, which this month unveiled measures to slash the 27-nation bloc’s dependence on America and Asia for key technologies, including Artificial Intelligence, said it was assessing the development. Gina Neff, Professor of Responsible Artificial Intelligence at Queen Mary University London, warned that restricting access could limit safe testing and international collaboration. She said the UK government’s Artificial Intelligence Security Institute found in its tests that the model could exploit defences and systems 73% of the time. Anthropic is also suing the Pentagon after being labelled a “supply chain risk”, while a US judge has ruled the directive cannot be enforced as the lawsuit continues.
