Anthropic Forced to Suspend Its AI Claude Fable 5 and Mythos 5

The company Anthropic was recently forced to take its artificial intelligence models, Claude Fable 5 and Mythos 5, offline in response to an export directive issued by the U.S. government. This directive prohibits access to these services for any foreign nationals. Since last Friday, Anthropic has been in discussions with the White House in an attempt to reach an agreement that would allow these offerings to be restored, but no compromise has yet been found.

Since the launch of Mythos in April, Anthropic has warned about the advanced capabilities of this model. Indeed, Mythos is capable of detecting software vulnerabilities, which can be useful for defenders seeking to patch these flaws. However, these same capabilities can also be exploited to find ways to take advantage of them, posing a potential risk if they fall into the wrong hands. Anthropic highlighted this risk during the launch of Mythos 5 and Claude Fable 5, stating that the advanced use of these AI models presents a dual-use dilemma.

To mitigate these risks, Anthropic initially released a version called Mythos Preview to a limited group as part of a project known as Project Glasswing. Last week, Mythos 5 was made available to this group privately, while Claude Fable 5, which is a model of similar capability to Mythos, was made accessible to the general public, but with specific restrictions regarding issues related to biology and cybersecurity.