Anthropic's recently unveiled cybersecurity tool, Mythos, has reportedly been accessed by a group of unauthorized users.
Mythos, an AI-powered product engineered for enterprise security, has been widely discussed for its potential dual nature: a powerful defense mechanism that, in the wrong hands, could become a formidable hacking instrument, as acknowledged by the company itself. Bloomberg has now reported that an undisclosed "private online forum" successfully gained entry to the tool through a third-party vendor.
An Anthropic spokesperson informed TechCrunch, “We’re investigating a report claiming unauthorized access to Claude Mythos Preview through one of our third-party vendor environments.” The company clarified that, to date, no evidence suggests this alleged unauthorized activity has impacted Anthropic's internal systems.
The group of unauthorized individuals reportedly employed various methods to access the model, including leveraging "access" held by a person interviewed by Bloomberg. This individual is currently employed by a third-party contractor working with Anthropic, according to the report.
Members of this group are associated with a Discord channel dedicated to discovering unreleased AI models, the outlet noted. They have reportedly been using Mythos consistently since gaining access and provided Bloomberg with supporting evidence, including screenshots and a live demonstration of the software.
Bloomberg's report indicates that the group, which purportedly obtained access on the very day Mythos was publicly announced, "made an educated guess about the model’s online location based on knowledge about the format Anthropic has used for other models." A source conveyed to the outlet that the group's primary interest lies in "playing around with new models, not wreaking havoc with them."
Mythos was initially distributed to a select cohort of vendors, including major entities like Apple, under an initiative known as Project Glasswing. This restricted release strategy was specifically designed to prevent its exploitation by malicious actors. Anthropic had previously warned that the tool possessed the capability to be weaponized against corporate security rather than bolstering it.
Should these claims of unauthorized use prove true, it could present significant challenges for Anthropic, particularly given that the exclusive release was intended to mitigate the company's concerns regarding enterprise security.
The Editorial Staff at AIChief is a team of professional content writers with extensive experience in AI and marketing. Founded in 2025, AIChief has quickly grown into the largest free AI resource hub in the industry.