Anthropic confirmed that hackers are accessing Claude user accounts by stealing login session cookies. This unauthorized access allows third parties to consume the users' subscription quotas without their knowledge. The company stated that attackers use common information-stealing malware to compromise user computers and extract session data.
Grant Desvott, an AI consultant, reported that his Claude Max 20x account, which costs $200 per month, showed abnormal usage even when inactive. Despite disabling all connected tools and cloud tasks, his usage increased from 45% to 55%. Anthropic eventually suspended his account, revoked all session tokens, and refunded £44.49 for the remaining subscription period.
The investigation revealed that a leaked Claude session key was used to generate unauthorized OAuth tokens for Claude Code. Anthropic noted that it cannot provide users with itemized usage logs, as their customer support system only tracks total consumption. This limitation makes it difficult for users to detect unauthorized activity promptly.
Multiple users on platforms like Reddit and GitHub have reported similar issues, including sudden account upgrades, unauthorized credit card charges, and rapid depletion of daily quotas. One user reported their usage jumping from 0% to 49% in just 12 minutes. Anthropic maintains that the malware is not spread through its own services but via external malicious downloads or ads.
Following the incident, Desvott canceled his subscription and switched to Cursor, citing the lack of tools to track usage sources and the slow response time from Anthropic support. When asked by TechCrunch how users can identify potential account abuse, Anthropic declined to comment.
