Back Theregister

Anthropic cracks down on hijacked user accounts mining AI tokens

Theregister • August 31, 2026

Share: View original

Commodity malware steals authenticated sessions, letting thieves freeload on victims' paid usage

security Researcher shows how Claude Code can be tricked simply by asking it to summarize a website

Researcher shows how Claude Code can be tricked simply by asking it to summarize a website

science German-Japanese researchers invent electricity-free tech that could cool datacenters

German-Japanese researchers invent electricity-free tech that could cool datacenters

Legal Nuisance-call blocker fined £190k for being a nuisance caller

Nuisance-call blocker fined £190k for being a nuisance caller

NETWORKS A lot of datacenter networks are run by absolute clowns. Not Amazon's

A lot of datacenter networks are run by absolute clowns. Not Amazon's

security Security vets rally around $4 paper password books for sale in Australia

Security vets rally around $4 paper password books for sale in Australia

Rather than paying for their own Claude usage, crims are using malware to steal access to other people's accounts. Aware of this issue, Anthropic has signed at least one affected user out and removed the saved payment method to stop stolen sessions being abused.

According to an email shared by user WorriedAssociate7029, who sent a copy to The Register , Anthropic has been keeping an eye on a threat actor using infostealer malware to hijack Claude login details, session cookies, and other info needed to subvert multifactor authentication on user accounts. Once obtained, the miscreant is using the stolen information to use premium Claude services without having to pay the bill themselves.

Fortunately for WorriedAssociate7029, Anthropic logged the user out of their account and deleted their stored payment method because it had detected evidence of attempted fraud.

“A few days ago, my social media accounts were hacked,” WorriedAssociate said, adding that they'd managed to track the malware down with the help of Claude Opus 5 Max and, they believe, cleaned the system. “But last night I received this email from Anthropic warning me of an attempt to steal tokens via the API.”

They explained that the attempt failed, apparently thanks to Anthropic spotting it, but they realized that meant that the cybercriminal behind the incident seemed to have hijacked Google account credentials, cookies, and session IDs as well, since that’s how they were signed into Claude. After changing their password again and removing all active sessions, it appears they are now safe.

Who’s eating your cookies?

Anthropic made clear in the email that the credential theft wave it’s identified has nothing to do with Claude itself, nor is it some sort of fancy, new-fangled, agentic AI malware that’s being used to create a base of accounts for bad actors to abuse. This is just good old-fashioned infostealer malware being turned to a new purpose, the email explains.

“We have no reason to believe that this malware is related to Claude, installed through Claude, or related to anything you did with Claude,” the email forwarded to us by WorriedAssociate and posted to stated. “Your Claude session was likely one of the many things it collected. It appears that a bad actor has now started picking the Claude sessions out of what it collected and using them.”

In this case, it’s well-known infostealing malware too: Vidar , LummaC2 , StealC , RedLine , Acreed, and Atomic Stealer have all been fingered by Anthropic as being used to steal Claude credentials, sessions, and cookies.

As for WorriedAssociate, they copped to making a noob mistake that led to their infection.

“I got fooled like a rookie by downloading a cracked game,” they admitted in a on their post. “Never again.”

As in their post, WorriedAssociate told us in a chat that they gave credit to Anthropic for cluing them in to the fact that they hadn’t fully secured their accounts, and said they appreciated what the company did to help lock their Claude account down.

“There have been several cases on in the past of accounts being hacked to steal tokens, and Anthropic’s customer service seems pretty dreadful when it comes to refunds and account recovery,” they told us. “This email appears to be new, and measures have finally been put in place to protect AI users.”

“Tokens are valuable and can be resold,” WorriedAssociate added. So let this be a lesson: Providers might not catch every case of account theft, and AI accounts are the new hotness. Don’t let your tokens be burned by someone else - they’re expensive and the last thing you want them to be used for is someone else's work. ®

Anthropic cracks down on hijacked user accounts mining AI tokens

Commodity malware steals authenticated sessions, letting thieves freeload on victims' paid usage

From watches to passwords on paper, everything that's old is new again

This week on the Kettle, we reminisce the good old days of Windows on XP's 25th anniversary, how new-old Casio watches could reshape the smartwatch market, and why passwords might be better on paper

Platform Engineering 2.0: your platform was built for a different era. AI just exposed it

PARTNER CONTENT: Platform engineering won the argument. Now it has to grow up fast and evolve for the AI era.

Energy biz SSE smacked around in court by a guy and AI

Company's three year pursuit of debt from non-existent address ended by Oxford judge

A lot of datacenter networks are run by absolute clowns. Not Amazon's

AWS keeps customer costs down in part through networking advances

The balkanization of virtualization will de-throne VMware, which doesn't mind a bit

The real fight is who wins the most AI and K8s workloads