Claude's Tokens Exploited by Hackers in Unauthorized Usage
Unauthorized access to user accounts on Anthropic's AI service "Claude" has been revealed after users noticed unexplained token consumption. Since last month, Anthropic has been issuing warnings to users about potential hacker threats.

Anthropic, which provides the AI service "Claude," has issued a warning regarding unauthorized access to user accounts. A user discovered that tokens—units representing AI processing volume—were being consumed on their account despite not using the service, which brought this issue to light. Anthropic subsequently issued an official notification to users about the possibility of hacker-related damages.
Claude operates on a subscription model where users pay a monthly fee to receive a certain amount of tokens. Since tokens are consumed through interactions with AI, the situation where tokens decrease despite no active usage suggests the possibility that a third party is illegally using the account. In such services, cases of account credential theft and misuse by others have become an increasingly noted concern as AI services proliferate.
The confirmed facts show that a Claude user discovered tokens disappearing even during times when they were not using the service last month (approximately one month before the report). Following this report, Anthropic issued user alerts and warnings suggesting hacker involvement. At present, details regarding the scale of damage or specific attack methods have not been confirmed.
This issue reveals that AI service accounts, like other online services, face the risk of unauthorized access. Particularly in token-based billing models, account takeover can result in unexpected charges or complete depletion of usage limits. Users are advised to regularly review their account usage history and check for any unrecognized token consumption.
With the expansion of AI services, there is growing expectation that attacks targeting authentication credentials and access tokens will increase. On the corporate side, companies are entering a phase where continuous security reinforcement is required, including mechanisms to detect abnormal access patterns and implementation of two-factor authentication. How Anthropic addresses this with concrete measures going forward will be closely watched.
This article is an original work independently written and edited by the AI issue editorial team based on factual reporting. © AI issue. Unauthorized reproduction, redistribution, or use for AI training is prohibited.