A South Korean developer under the pseudonym remy_notes fell victim to a glaring error in the billing system of the AI startup Anthropic. While on the free Claude plan, he received an astronomical bill of $16.6 million. The incident, which he reported on the social network Threads, began on July 7 with an initial invoice for $1.6 million.
Initially, the programmer suspected phishing, but the next day a second payment demand arrived—this time for $16.6 million. The links and sender data led to Anthropic's official domain and its payment system, Stripe. Meanwhile, remy_notes' dashboard showed zero API usage, and there were no billable keys on the account.
The situation worsened on July 8, when the bank twice declined withdrawal attempts exceeding the per-transaction limit. The payments came from a U.S. merchant named ANTHROP. The user thoroughly checked all processes on his Mac mini, including eight AI agents, but found nothing that could explain the glitch. As a preventive measure, he canceled his Claude Max subscription and blocked his card.
It took four days and about 18 emails to support to clarify the circumstances. Some responses came automatically—from Anthropic's AI agent named Fin. Only on July 12 did the company confirm the error, citing a misconfigured auto-recharge balance feature as the cause. Anthropic stated that it had not charged a single dollar and disabled this feature as a precaution. They linked the developer's card block to failed withdrawal attempts, emphasizing that the account had not been hacked. However, support could not explain why the system generated exactly $16.6 million.
Not an Isolated Case: Systemic Billing Errors
This incident is just the tip of the iceberg. Earlier, the auditing firm Vaudit identified massive overpayments by corporate clients of Anthropic and OpenAI. From March to June, they reviewed invoices totaling $34 million across 60 organizations. Most invoices were related to the use of Anthropic's Claude Code. Specialists discovered excessive charges amounting to about $1.7 million—roughly 5% of the audited volume.
Vaudit highlighted several systemic errors. Clients were billed at the rates of new models, even though a cheaper older version was actually used. Charges were recorded for requests that ended in errors or produced no results. A separate category consisted of "retry storms"—when an AI agent repeatedly retries a failed task in the background, and each attempt appears on the bill. After Vaudit and clients filed complaints, companies began refunding funds—about 80% of the overpayments were compensated by Amazon, Google, Microsoft, Anthropic, and OpenAI.
My comment as an analyst: Such errors undermine trust in the infrastructure of AI services. Auto-payment and billing systems in these companies must be flawless, especially when dealing with sums comparable to the GDP of a small country. The incident with remy_notes is not just a curiosity but a signal that Anthropic's internal cost control algorithms require serious improvement. If a company cannot explain the generation of a $16.6 million bill, then its billing system is a "black box" that could strike any user.