Anthropic has begun notifying Claude users that infostealer malware on their personal computers has stolen active login sessions, allowing attackers to consume their usage limits without permission. The company is signing out affected users, wiping saved payment methods, and refunding unauthorized charges.
The malware families identified include Vidar, LummaC2, StealC, RedLine, and Acreed on Windows, and Atomic Stealer (AMOS) on a smaller number of Macs. Anthropic traced the infections to pirated software downloads and malicious applications rather than any vulnerability in Claude itself.
The tell for affected users, Anthropic said, is that "usage limits looked like they refilled and then drained while you weren't using Claude." The stolen session tokens allow attackers to use the victim's Claude subscription as if they were the legitimate user, running up usage on the victim's account.
The incident highlights a growing security challenge for AI services: as subscription-based AI tools become more valuable, they become increasingly attractive targets for credential-stealing malware. Session hijacking through infostealers is a well-understood attack vector in traditional web services, but the high per-use cost of AI inference makes stolen AI sessions particularly lucrative for attackers.
Anthropic emphasized that no breach of its own systems occurred. The malware operates by stealing browser cookies and session tokens from infected machines, a technique that bypasses even strong account passwords. Users are advised to ensure their operating systems and browsers are up to date and to avoid downloading software from untrusted sources.
