The US Cybersecurity and Infrastructure Safety Company (CISA) doesn’t but have entry to Anthropic’s bug-hunting AI mannequin, Claude Mythos, though different authorities businesses do, Axios reported earlier this week.
As if that weren’t a large enough slap within the face for the nationwide cyber-defense company, the listing of those that do have entry to Mythos consists of a number of unauthorized customers, in line with Bloomberg Information. Members of a non-public Discord channel specializing in in search of details about unreleased AI fashions, have gained entry to Mythos, in line with one unnamed member of the group, Bloomberg reported. “The group has been utilizing Mythos usually since then, although not for cybersecurity functions,” the particular person advised Bloomberg, supplying screenshots to again up their declare.
Because of its concern that the highly effective mannequin may very well be used to determine and exploit flaws in software program and on-line companies, Anthropic has restricted entry to a preview of Mythos to an unique group of presidency businesses, trade teams, and software program suppliers by way of an initiative it calls Undertaking Glasswing.
