Anthropic accidentally billed a developer $16.6 million — then took 4 days to reply

What’s worse than a $16.6 million bill you never asked for? The four days of silence that follow when you try to get an explanation.

A Korean developer known as @remy_notes found out the hard way last week. On July 11, his bank account was hit with two charges from Anthropic: one for $1.6 million, another for $16.62 million. The surreal part — when he checked his Claude API console, it showed a balance of exactly $0. No activity. No charges. Nothing.

Anthropic has now confirmed the cause: a billing system bug. No actual money was deducted, the company says. But that’s only half the story. While the charges never went through, Anthropic’s system had already sent multiple payment attempts to the developer’s bank. Enough to trip fraud detection and freeze his credit card entirely.

Then came the customer service ordeal. Over four days, @remy_notes sent 15 emails across Anthropic’s customer support, technical, and marketing channels. The count eventually climbed to 18. What did he get back? On day four, an automated notification saying the issue had been escalated to “the right person.” When that turned out to be another auto-reply, the developer could only joke that being “upgraded to human processing” apparently means getting a slightly different automated email.

As AI companies race to sign up enterprise customers and push toward billion-dollar revenues, it’s the unglamorous plumbing — billing infrastructure, account management, customer support — that can break in spectacular ways. A developer with a frozen credit card and an inbox full of auto-replies is left wondering whether anyone is actually home.