Warning: Hackers Target Claude Subscribers to Steal Tokens
On August 4, Grant De Swardt, a freelance AI consultant based in East Sussex, U.K., noticed unusual activity with his Claude Max 20x account. Despite not working that day, he observed an increase in token usage. The following day, after disabling all integrations and not performing any tasks, his token consumption still rose—confirming something was amiss.
Curious and concerned, he reached out to Anthropic for an explanation. Although they did not provide an itemized usage report, they acknowledged a problem. As a precaution, Anthropic suspended his account, invalidated all sessions, and issued a partial refund of £44.49 for his subscription, which costs $200 per month.
This suspension disrupted De Swardt’s business, which relied heavily on AI agents for tasks such as inputting purchase-order data into accounting systems and managing various administrative functions. After their investigation, Anthropic informed De Swardt that a compromised Claude session key had been used to generate unauthorized tokens. The unauthorized use had likely involved a third-party service, but Anthropic could not determine how the access was gained.
De Swardt shared his experience on Reddit and discovered he was not alone; others had reported similar issues, with tokens being consumed without their direct interaction with the service. Some users detailed extreme spikes in usage during periods of inactivity, leading to concerns about unauthorized access.
In emails from Anthropic to other affected users, the company warned that a bad actor was leveraging infostealer malware to obtain login sessions from users’ computers. This type of malware can pilfer saved credentials through various means, which does not include using Claude directly. While some users received notifications about suspicious activity and were signed out or had their authorizations canceled, De Swardt did not receive such an alert, raising further concerns for him.
After a two-week suspension, De Swardt’s account was reinstated, but the lack of prompt assistance and transparency regarding usage issues led him to cancel his subscription. He switched to a competing service called Cursor, which provided access to multiple models, including more affordable open-source options. De Swardt noted that he found similar functionality in other models and could not justify returning to Claude unless there were significant improvements in security and transparency.
Despite his advocacy for better user tools to monitor token usage, Anthropic declined to comment on measures for identifying misuse, leaving concerns about user protection unresolved.
Discover the pinnacle of WordPress auto blogging technology with AutomationTools.AI. Harnessing the power of cutting-edge AI algorithms, AutomationTools.AI emerges as the foremost solution for effortlessly curating content from RSS feeds directly to your WordPress platform. Say goodbye to manual content curation and hello to seamless automation, as this innovative tool streamlines the process, saving you time and effort. Stay ahead of the curve in content management and elevate your WordPress website with AutomationTools.AI—the ultimate choice for efficient, dynamic, and hassle-free auto blogging. Learn More
