Hackers used stolen Claude sessions to consume subscribers’ tokens
Grant De Swardt, an independent AI consultant in East Sussex, U.K., noticed on August 4 that tokens were being consumed from his Claude Max 20x account even though he was not working. After disabling connected tools and pausing scheduled tasks, usage still rose from 45% to 55%. Anthropic suspended the account, invalidated its sessions and server-side Claude Code tokens, and provided a partial refund of £44.49 from his $200-per-month subscription.
Anthropic later told De Swardt that a compromised Claude session key had been used to create unauthorized Claude Code OAuth tokens. The company said the account appeared to have supported activity for other users, but it could not establish how access was obtained. De Swardt said he found no evidence that his computer had been compromised.
Other Claude users reported unexplained usage increases on Reddit and GitHub. Anthropic told some users that infostealer malware had taken login sessions from their computers, allowing accounts to be accessed and tokens consumed. The company signed out affected users, invalidated authorizations, issued some refunds and warned that malware might be present. De Swardt’s account was restored after about two weeks, but he later cancelled his subscription for Cursor, citing slow support and the absence of itemized usage information.
Written by Sakhanda Wire AI from reporting by TechCrunch. Facts, quotes and full context are in the original article below.