In early July, Anthropic, the developer of the well-known AI model Claude, became the epicenter of an unusual incident that exposed serious vulnerabilities in the billing systems of AI giants. A user from South Korea, going by the nickname remy_notes, received a bill for more than $16.6 million, despite not using paid APIs and not even linking a bank card to the service.
The situation developed rapidly. The first bill for $1.67 million arrived on July 7. By the next day, the amount had increased tenfold—to $16.6 million. Moreover, the developer, who used exclusively the free tariff plan, had taken no actions that could have led to such charges. In the personal account, API consumption remained at zero, payment keys were not activated, and no card was linked.
How the bank and support responded to the error
Attempts to debit the amount were rejected by the bank, as the sums exceeded the card limit. However, after several failed transactions, the bank blocked the user's card. Resolving the issue with Anthropic took four days and about 18 emails to customer support. Instead of a coherent response, the developer received automated replies. Only after an ultimatum with a deadline set for Monday Korean time did the company acknowledge the error.
Anthropic stated that the incident occurred due to a malfunction in the settings of the auto-recharge credit feature. The company disabled this option "for security reasons" but did not explain why the auto-recharge amount reached such absurd values. It is important to note that the money was not actually debited—all transactions were rejected by the bank.
A systemic industry problem
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 occurred precisely in AI solution billing. This indicates that the problem is systemic: the billing systems of many AI providers are simply not ready for the scale and complexity of modern calculations.
My analysis: The AI services market is growing explosively, but the billing infrastructure is clearly lagging behind. Such failures are not just annoying errors but a signal for the entire industry. Developers and users should regularly check payment panels and monitor bank notifications. A random glitch can lead to serious financial problems if not addressed in time.