Anthropic billing glitch: South Korean user receives 'phantom' bill for $16.6 million
An incident that occurred in early July with a user from South Korea exposed serious vulnerabilities in the payment systems of companies operating in the field of artificial intelligence. A developer under the pseudonym remy_notes, using Anthropic's free plan, discovered two astronomical bills in their account — $1.67 million and, less than a day later, $16.6 million. Notably, the user had not made any API transactions, linked a payment card, or generated paid keys.
How the system "charged" millions
The first notification arrived on July 7. Suspecting phishing, the user verified the authenticity of the emails: they came from Anthropic's official domain and through the real payment service Stripe. In the personal account, API consumption remained at zero, and scripts and AI agents showed no anomalies. The bank rejected two debit attempts on July 8 due to exceeding the card limit, but after several repeated requests, the card was eventually blocked.
Anthropic's response and the scale of the problem
The company acknowledged the error, explaining it as a glitch in the settings of the auto-replenishment function. The function was disabled "for security reasons," but Anthropic could not clearly explain why the auto-replenishment amount reached such absurd values. Resolving the issue took four days and about 18 emails — instead of a substantive response, the user received only automatic replies. They were forced to set a deadline for the company by the end of Monday Korean time.
Systemic failure in the industry
This case is not isolated. Startup Vaudit audited $34 million in AI service bills for 60 major clients and identified overpayments of approximately $1.7 million (about 5% of the total amount). The largest errors are attributed to billing for AI-based solutions, and among the affected are giants such as Panasonic, HP, and Honda.
Expert opinion: This incident is a warning signal for the entire industry. Billing problems in the AI sector are becoming systemic, and they impact not only retail users but also corporate clients. Until companies like Anthropic implement strict verification and transaction limit mechanisms, users will have to rely solely on their own vigilance: regularly checking payment panels, monitoring bank notifications, and promptly responding to any suspicious charges. Otherwise, a random glitch could turn into a serious financial problem.