On August 4, Grant de Swardt, an independent AI consultant based in East Sussex, UK, noticed unusual activity on his Claude Max 20x account. He observed that his token usage — a measure of how much AI processing power he was using — was increasing even though he wasn’t working that day. The next day, he disabled all connected services, but the token consumption continued to rise. During a controlled test, de Swardt found that usage jumped from 45% to 55% without any active work, scheduled tasks, or cloud-based processes running. When he contacted Anthropic, the company behind Claude, he was unable to get an itemized breakdown of his usage. In response, Anthropic suspended his paid account, revoked all active sessions and tokens, and issued a partial refund of £44.49 for the remaining time on his $200-per-month subscription. This disruption affected his business, which relies on AI tools to automate tasks such as extracting purchase-order data from emails and inputting it into accounting software. Anthropic later explained that the issue stemmed from a compromised Claude session key, which was used to generate unauthorized OAuth tokens. The company stated that the account "appeared to have been used by an unauthorized-looking third-party service to handle activity for other people." However, it could not determine how the access was initially obtained. De Swardt noted that Anthropic could not confirm whether his credentials were stolen without his knowledge or if the account had been linked to an external service. After sharing his experience on Reddit, de Swardt discovered that he was not alone. Other users reported similar issues, including accounts being auto-upgraded without their consent, leading to unexpected charges and full token usage without any interaction. One user saw their token usage jump from 0% to 49% in just 12 minutes, while another experienced maximum token consumption for three consecutive days without using the service. Some users shared emails from Anthropic that warned them about potential token theft by a third party using infostealer malware — a type of malicious software that steals login credentials and session data from a user's computer. Anthropic advised affected users to check their systems for malware, though it clarified that the malware did not originate from Claude itself. De Swardt found no evidence that his own computer had been compromised and remains uncertain about how hackers gained access. His account was eventually reinstated after two weeks, but the lack of detailed usage reports and the slow response from Anthropic led him to cancel his subscription. He now uses Cursor, an alternative AI platform that supports multiple models, including more affordable open-source options. He noted that these models perform similarly to Claude and expressed doubt about returning to Anthropic unless the issue is fully resolved. Anthropic did not provide specific tools or guidance for users to detect or prevent such misuse when asked for more information.