A gaggle of unauthorized customers has reportedly gained entry to Mythos, the cybersecurity software lately introduced by Anthropic.
A lot has been product of Mythos and its purported energy — an AI product designed for enterprise safety that, within the incorrect palms, may grow to be a potent hacking software, in keeping with the corporate. Now, Bloomberg has reported {that a} “non-public on-line discussion board,” the members of which haven’t been publicly recognized, has managed to realize entry to the software by a third-party vendor.
“We’re investigating a report claiming unauthorized entry to Claude Mythos Preview by one in every of our third-party vendor environments,” an Anthropic spokesperson advised TechCrunch. The corporate mentioned that, up to now, it has discovered no proof that the supposedly unauthorized exercise impacted Anthropic’s techniques in any respect.
The unauthorized group tried a variety of totally different methods to realize entry to the mannequin, together with utilizing “entry” loved by the one that was interviewed by Bloomberg. That individual is presently employed at a third-party contractor that works for Anthropic, the outlet reported.
Members of the group are a part of a Discord channel that search out details about unreleased AI fashions, the outlet reported. The group has been utilizing Mythos recurrently since having access to it, and supplied proof to Bloomberg within the type of screenshots and a dwell demonstration of the software program.
Bloomberg reviews that the group, which supposedly gained entry to the software on the exact same day it was publicly introduced, “made an informed guess concerning the mannequin’s on-line location based mostly on data concerning the format Anthropic has used for different fashions.” The group in query is “desirous about taking part in round with new fashions, not wreaking havoc with them,” the supply advised the outlet.
Mythos was launched to a choose variety of distributors, together with massive names like Apple, as a part of an initiative referred to as Undertaking Glasswing. The restricted launch of the mannequin was designed to cease its utilization by dangerous actors. The software may very well be weaponized towards company safety as a substitute of bolstering it, Anthropic mentioned.
Techcrunch occasion
San Francisco, CA
|
October 13-15, 2026
If true, unauthorized use of Mythos may spell hassle for Anthropic, which supplied the unique launch to allay the corporate’s concern for enterprise safety.
Whenever you buy by hyperlinks in our articles, we may earn a small commission. This doesn’t have an effect on our editorial independence.
