Hackers are stealing Claude tokens from subscribers

For many AI power users, Claude has become an indispensable engine for business operations. However, a growing number of subscribers are discovering that their accounts are being weaponized by bad actors to siphon off expensive token allowances. The issue came to light following reports of mysterious, unauthorized activity that left users footing the bill for work they never performed.

The Mystery of the Vanishing Tokens

The alarm was first raised in early August by Grant de Swardt, an independent AI consultant based in the UK. Despite having no active tasks running and his automated workflows paused, de Swardt noticed his Claude Max 20x account was rapidly consuming its token limit.

"In the clearest controlled interval, it increased from 45% to 55% while I performed no work," de Swardt explained. After disabling all integrations and local tasks, the consumption continued unabated. When he reached out to Anthropic for an itemized breakdown of his usage, the company was unable to provide one, though they acknowledged that the activity was indeed anomalous.

Anthropic eventually suspended his account, invalidated his active sessions, and issued a partial refund of £44.49. However, the disruption caused significant friction for de Swardt, whose business relies heavily on AI-driven agents for administrative and coding tasks.

A Widespread Security Concern

Upon investigating the breach, Anthropic informed de Swardt that a compromised session key had been used to mint unauthorized OAuth tokens. The company suggested that his account was likely being leveraged by a third-party service to process requests for other users, though the exact entry point for the hackers remained unclear.

De Swardt’s experience is far from an isolated incident. A subsequent investigation into community forums like Reddit and GitHub revealed a pattern of similar complaints:

• Unauthorized Upgrades: Users reported accounts being upgraded to higher tiers without consent, with credit cards charged automatically.
• Rapid Depletion: Reports of accounts hitting 100% usage within minutes of inactivity.
• Persistent Drain: Some users noted their daily token limits were being exhausted consistently over several days, even while they were logged out.

"We have 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 stated in correspondence with affected users.

The Role of Infostealers

Anthropic’s internal findings point toward infostealer malware as the primary culprit. These malicious programs surreptitiously install themselves on a user’s machine to scrape saved passwords, browser cookies, and active session tokens. By hijacking these session files, hackers can bypass traditional login requirements, effectively "becoming" the user without needing a password or multi-factor authentication.

The company has been proactive in some cases, signing out suspicious sessions and warning users that their local machines may be infected. Crucially, Anthropic clarified that the malware is not being distributed through their own platform, but rather through common vectors like malicious advertisements or infected software downloads.

The Fallout and Future Outlook

Despite the company’s efforts to mitigate the damage, the lack of transparency regarding usage logs remains a major pain point for subscribers. De Swardt, who had his account reinstated after two weeks, ultimately decided to migrate his workflow to Cursor. He cited the platform's ability to utilize multiple models—including more cost-effective open-source alternatives—as a key advantage.

For de Swardt, the experience highlighted a fundamental gap in AI platform security: the inability for users to audit their own consumption. "I don’t think there’s any way that these people can protect themselves," he noted, expressing frustration that Anthropic still lacks granular tools for users to monitor exactly what is consuming their tokens.

Key Takeaways for Users:

• Monitor Usage: Keep a close eye on your account dashboard for spikes in activity during periods of inactivity.
• Secure Your Machine: Since session hijacking is the primary vector, ensure your computer is free of malware and avoid downloading suspicious files.
• Demand Transparency: If you notice anomalous activity, contact support immediately and request a formal review of your session logs.
• Consider Alternatives: If your business relies on high-volume token usage, consider platforms that offer better visibility and multi-model flexibility.

When approached for further comment on how users might better identify and prevent such misuse, Anthropic declined to provide additional details. As the AI landscape continues to evolve, the burden of security remains a shared responsibility, yet users are clearly demanding more robust tools to safeguard their accounts from these sophisticated digital thefts.