Anthropic Warns Infostealer Malware Is Hijacking Claude Sessions to Drain Usage

My take: For those of us who use AI tools in our daily work, this is a warning worth paying attention to: infostealer malware is now specifically targeting Claude, with attackers using stolen sessions to consume usage credits without permission. Anthropic identified six distinct malware families in this campaign, including Vidar, LummaC2, and AMOS on macOS, all designed to steal authentication cookies stored in browsers.

The good news is that Anthropic is acting quickly: it is signing out compromised sessions, removing saved payment methods, and refunding unauthorized charges. But this incident makes clear that protecting your AI tools requires the same security controls you would apply to your bank account or corporate email.

If you use Claude, check whether you received a notification about compromised sessions. And if you are a business owner or manager of a team that works with AI tools, do you already have clear policies on device security and credential management for those tools?