Hackers are stealing Claude tokens from subscribers, users say

Grant De Swardt, an independent AI consultant in East Sussex, U.K., noticed on August 4 that his Claude Max 20x account’s token usage was climbing even though he hadn’t been working that day. The next day he disabled everything connected to Claude and still didn’t use it — yet consumption kept rising. “In the clearest controlled interval, it increased from 45% to 55% while I performed no work, scheduled Cowork tasks were paused or completed, Dispatch/cloud execution was disabled, and there was no corresponding active local Claude Code task,” De Swardt told TechCrunch.

He contacted Anthropic and asked for an itemized usage list, which the company didn’t provide, though it agreed something was off: it suspended his paid account, invalidated his sessions and server-side Claude Code tokens, and issued a partial refund of £44.49 on his $200-per-month subscription. Anthropic later told De Swardt it had found the cause — a compromised Claude session key used to mint unauthorized Claude Code OAuth tokens — and that the account “appeared to have been used by an unauthorized-looking third-party service to handle activity for other people,” though it couldn’t determine how access was obtained; Anthropic said the evidence was consistent with either credentials being taken without his knowledge or the account being connected to an outside service.

De Swardt, who relies on Claude and other AI tools throughout his business helping small and mid-size companies set up agents, posted his experience on Reddit and found he wasn’t alone: other users reported accounts auto-upgraded without consent, usage jumping from near zero to full in minutes, or tokens burning through daily limits for days without any use. Two users shared emails in which Anthropic had proactively identified and warned them of the theft, saying it had “recently become aware of a bad actor that is using common infostealer malware to steal Claude login sessions from people’s computers, then using those login sessions to access Claude accounts and consume their usage.” Anthropic said the malware doesn’t come from using Claude itself, and that when it spots suspicious activity it signs users out, invalidates authorizations, and issues refunds.

De Swardt did not receive one of those proactive warning emails and says he found no evidence his own computer was compromised. His account was reinstated after about two weeks, but the slow response and the lack of itemized usage data pushed him to cancel his subscription in favor of a multi-model alternative; Anthropic declined to comment when asked how users can identify misuse.

Why it matters

Without itemized usage tracking, Anthropic’s subscribers have no way to independently verify or catch token theft themselves, leaving detection dependent entirely on the company noticing anomalies first.