On August 4, Grant De Swardt, an AI consultant in East Sussex, U.K., observed unusual token consumption on his Claude Max 20x account despite not actively using it. The following day, even with all connected services disabled, his token usage continued to rise. After contacting Anthropic, De Swardt was informed that a compromised Claude session key had been exploited to generate unauthorized Claude Code OAuth tokens, leading to the suspension of his paid account and a partial refund of £44.49. Anthropic determined that the account had been accessed by an unauthorized third-party service, though the exact method of access remained undetermined.
De Swardt’s experience, shared on Reddit, resonated with other users who reported similar incidents of unauthorized token usage, including automatic account upgrades and sudden spikes in token consumption. Anthropic later disclosed that a “bad actor” was employing infostealer malware to pilfer Claude login sessions from users’ computers, subsequently accessing and depleting their token allowances. The company advised affected users of potential malware infections and took measures such as invalidating authorizations and issuing refunds. However, De Swardt, who found no evidence of malware on his system, criticized Anthropic for the lack of itemized usage data and the difficulty in obtaining timely support, ultimately canceling his subscription in favor of alternative models. Anthropic declined to provide guidance on identifying token misuse. Source