In early July 2025, Anthropic, the developer of advanced AI models, faced a serious billing system failure. The error resulted in a user from South Korea, who was on a free plan, being billed an astronomical amount—over $16.6 million. This incident, which exposed systemic issues in AI service billing, deserves close attention.
A developer under the pseudonym remy_notes, who had not linked a payment card to their account and did not use the API, began receiving notifications from Anthropic. The first invoice for $1.67 million arrived on July 7. By the next day, the amount had increased tenfold—to $16,627,739. All notifications came from Anthropic's official domain via the Stripe payment service, ruling out the possibility of phishing.
How It Happened
In the user's personal account, there were no API requests and no paid keys created. The developer thoroughly checked their scripts and AI agents but found no reason for the charges. The problem worsened when the bank, after twice declining debit attempts due to card limit exceedance, eventually blocked the card. It took support four days and about 18 emails to resolve the situation.
Official Position and Hidden Risks
Anthropic acknowledged the error, stating that no money was debited. The cause was attributed to a failure in the auto-recharge credit settings, which have since been disabled. However, the company did not explain how the auto-recharge amount could reach such absurd values. This raises questions about the reliability of internal control systems.
This case is not isolated. Vaudit, a startup specializing in AI expense auditing, reviewed $34 million in invoices for 60 major clients, including Panasonic, HP, and Honda. The results showed overpayments of approximately $1.7 million, or about 5% of the total amount. The most critical errors occur precisely in AI solution billing.
My analysis: The AI services market is growing exponentially, but the billing infrastructure is clearly lagging behind. This incident is not just an unfortunate oversight but a signal that automatic billing systems require fundamentally different protection mechanisms. Users, especially corporate ones, should implement regular payment audits and set strict transaction limits. Otherwise, a random glitch could turn into a real financial catastrophe.