Just days after the launch of the artificial intelligence model Fable 5, Anthropic suspended its operation because the US authorities expressed safety concerns.
In a statement published on her website, Anthropic reported that she was instructed to ban foreigners—even foreigners living in the US—the use of Claude Fable.
Anthropic made the decision to deactivate access for Fable 5 and limited access Mythos 5 for all its customers.
In the meantime, like notes BBC, Anthropic and Trump government are involved in a separate judicial dispute over a mandate that prohibits government agencies from using the company's artificial intelligence tools.
In her statement Anthropic said the US National Security Authorities had not mentioned specific problems of Fable. The government believes, however, that it has been informed of a jailbreaking method of the model.
Jailbreaking is a process of bypassing restrictions to protect the model, allowing malicious hackers to unblock functions.
"We have examined a demonstration of this particular technique used to identify a small number of already known, minor vulnerabilitys. All these vulnerabilitys seem relatively simple and we found that other models available to the public were able to identify them," Anthropic said.
Prior to the release of Claude Fable 5, leading executives from economics, technology and the state sector had expressed concerns about the public disposal of the model. Anthropic allowed access before traffic to a few selected persons, because the tool was so powerful that it could be dangerous due to its ability to exploit or violate computer systems.