The AI startup Anthropic found itself at the center of a scandal after mistakenly issuing a bill for a colossal sixteen point six million dollars to a user on a free Claude tier. This incident was reported by a South Korean developer known by the pseudonym remy_notes on the Threads social network. On July 7, the programmer received an initial invoice for one point six million dollars, which he initially dismissed as a phishing attempt. However, less than a day later, a second payment demand arrived, with the amount skyrocketing to sixteen point six million dollars. Notably, both the sender and the payment link directed to the official Anthropic domain and its Stripe payment system, while the user's account dashboard showed zero API usage and no billable keys were present. After the developer's bank card twice declined the withdrawal attempts due to exceeding the single-transaction limit, he proactively canceled his Claude Max subscription and blocked the card. Resolving the issue took four days and approximately eighteen emails to customer support, with some responses automatically generated by an AI agent named Fin.

On July 12, Anthropic officially confirmed the error in a letter to the user, assuring him that not a single dollar had been withdrawn from his account. The company attributed the incident to a misconfigured auto-top-up balance feature, which was immediately disabled as a precautionary measure. Anthropic representatives also linked the subsequent blocking of the developer's bank card directly to these failed withdrawal attempts initiated by the American seller named ANTHROP. The project team emphasized that the programmer's account had not been compromised or hacked by any third parties. However, the support team was unable to explain how the internal system generated this specific astronomical figure of sixteen point six million dollars, leaving the technical root cause of the exact amount unaddressed.

This incident is not an isolated case within the rapidly evolving artificial intelligence industry. In late June, The Information, citing the auditing firm Vaudit, reported widespread overcharging affecting corporate clients of both Anthropic and OpenAI. Between March and June, auditors reviewed invoices from sixty organizations totaling thirty-four million dollars and identified excessive charges amounting to approximately one point seven million dollars, representing about five percent of the audited volume. Specialists highlighted several systemic billing errors: clients were sometimes billed at the rates of newer models even when older, cheaper versions were actually used, and charges were recorded for requests that ultimately failed or yielded no results. A separate category of issues involved so-called "storms of retries," a situation where an AI agent repeatedly attempts a failed task in the background, with each attempt erroneously reflected in the final invoice. Following complaints from Vaudit and affected clients, the involved tech giants began issuing refunds, successfully compensating approximately eighty percent of the overcharged amounts.