Anthropic, the developer of the popular Claude model, has found itself at the center of a scandal due to a severe billing system glitch. A user from South Korea, on a free plan, received an astronomical bill of $16.6 million. Notably, the developer under the pseudonym remy_notes had not only not used any paid API keys but had not even linked a bank card to their account.
The incident began on July 7, when the user's inbox received the first notification for $1.67 million. The very next day, the amount increased tenfold to $16.6 million. Throughout this time, the developer had not performed any actions in the system: their dashboard showed zero API usage, and no paid keys were present. Nevertheless, the emails came from Anthropic's official domain via the Stripe payment service, ruling out the possibility of phishing.
Bank's Reaction and Four Days of Correspondence
The user's bank twice declined attempts to debit the funds on July 8, as the amounts exceeded the card's limit. However, after several failed attempts, the main card was blocked. To obtain written confirmation of the invoice cancellation, the developer needed four days and about 18 emails to Anthropic's support service. Instead of a clear response, they received only automated replies.
Anthropic acknowledged the error, explaining it as a glitch in the auto-recharge credit feature settings. The company stated that the money was not actually debited, and the feature itself has been disabled to prevent similar incidents. However, Anthropic representatives could not clearly explain why the auto-recharge amount reached such absurd values. The user had to issue an ultimatum with a deadline of the end of Monday, Korean time.
Systemic Problem: Billing Errors in AI Companies
This case is not isolated. The startup Vaudit conducted an audit of $34 million in AI service bills for 60 major clients, including Panasonic, HP, and Honda. The results showed overpayments of approximately $1.7 million — about 5% of the total amount. The largest errors are specifically in the billing of AI-based solutions. This indicates that the problem is systemic and affects the entire industry.
My expert commentary: This incident is a vivid example of how automated billing systems in the AI field can malfunction with catastrophic consequences. The market urgently needs more transparent mechanisms for controlling and limiting expenses, especially for users on free plans. Until companies like Anthropic implement strict security and notification protocols, such "phantom bills" will undermine trust in the industry. Users, for their part, should make it a rule to regularly check payment panels and promptly respond to any suspicious notifications from the bank.