The alarm was first raised by Grant de Swardt, an independent AI consultant based in East Sussex, UK. De Swardt noticed a disturbing trend in his Claude Max 20x account: his token consumption was climbing rapidly even during periods of total inactivity. Despite disabling all connected agents and pausing scheduled tasks, his usage surged from 45% to 55% in a controlled interval where no work was performed.

Upon contacting Anthropic, the company took drastic measures by suspending his account and invalidating all server-side tokens. The investigation revealed that a compromised session key had been used to mint unauthorized OAuth tokens. Essentially, a third-party actor had gained covert access to his account to siphoning off his expensive monthly token allowance for other users.

Why This Matters

BozokMedia analysis shows that this is a systemic vulnerability rather than an isolated incident. As enterprises increasingly integrate AI agents into their core workflows—as de Swardt does for small businesses—the theft of tokens represents a direct financial hit and a potential security breach. The lack of itemized usage logs means such thefts can go undetected for months, leaving users blind to how their resources are being consumed.

Infostealer malware bypasses traditional security by stealing active session cookies, allowing attackers to impersonate users without ever needing a password or a 2FA code.

The scope of the problem expanded as de Swardt shared his experience on Reddit and GitHub. Numerous users reported similar anomalies, with some claiming their accounts were auto-upgraded without consent, leading to unauthorized credit card charges. Anthropic eventually confirmed to some users via email that 'bad actors' are using common infostealer malware to steal login sessions from personal computers.

Historical Background: Infostealers, such as RedLine or Vidar, have long been the bane of digital security, primarily targeting crypto wallets and banking credentials. The shift toward targeting AI tokens reflects the growing market value of LLM compute power. Attackers can steal these tokens and resell access to high-tier AI models on the dark web, effectively monetizing stolen subscriptions.

Frequently Asked Questions

Q1: Does using Claude AI cause the malware infection?
No, Anthropic stated the malware is picked up from external sources, such as downloading infected software or clicking on malicious ads.

Q2: How can I tell if my tokens are being stolen?
Currently, users must manually monitor their total usage percentage, as Anthropic does not yet provide an itemized list of token consumption.