Anthropic Charged a Dev $180 While He Was Out Sailing. Then Ignored Him for a Month.
Here’s a fun Friday story.
UPDATE (2026-04-13): After Nick’s post went viral on Hacker News, a human support agent from Anthropic finally reached out and the billing issue is being worked through. Lesson stays the same: it took a front-page post to get a reply.
A developer named Nick went sailing in San Diego in early March. While he was on the water — pleasantly *not* using Claude, because you know, he was on a boat — his Anthropic account quietly racked up sixteen separate “Extra Usage” invoices between March 3 and March 5. Each one $10 to $13. Total damage: about $180.
Nick is on the Claude Max plan. That’s the “unlimited except when we say so” tier we wrote about yesterday. He was not using it. His Claude Code session history shows two tiny sessions totalling under 7KB on March 5. Seven kilobytes. That’s a grocery list. And yet his dashboard showed 100% session utilization. Billed accordingly.
So he did what any reasonable person does: he emailed support with the evidence. Here’s what happened next.
Day 1 (March 7). Instant reply from Anthropic’s AI chatbot, pointing him to a refund flow that didn’t apply to his situation. Classic.
Day 1, five minutes later. Nick replies, politely asking for a human. Gets a generic *”a member of our team will be in touch soon”* acknowledgment.
Day 10, Day 18, Day 32. Three follow-ups. Complete silence. No human. No update. No refund. Nothing.
Day 33 (April 9). Nick publishes his complaint on his personal blog. It hits the Hacker News front page within hours. Then, and only then, a human from Anthropic finally shows up and fixes it — refunding $190.13 and, in the process, discovering that someone had been logged into Nick’s account from a second device since February.
Read that last part again. Nick had been getting billed for sessions from a device he didn’t own, for over a month, and Anthropic’s support system didn’t catch it, didn’t investigate it, and didn’t respond to him until his complaint went viral.
This isn’t one guy’s bad luck
I went into the Hacker News thread (433 points, 200 comments) expecting the usual “well, my experience was actually fine” contrarians. I did not find them.
Instead I found a parade of people saying the exact same thing.
– A Developer Platform user waited two weeks across three different cards (Mastercard, Visa, AmEx) while the chatbot insisted payments were failing and support refused to investigate.
– An organization hit a billing loop bug where they couldn’t cancel, couldn’t add users, couldn’t delete users. They *abandoned the account*. Gave up.
– Someone on the Max x5 plan got a surprise $45.08 invoice despite having auto top-up *disabled*. Still waiting for a follow-up.
– Enterprise customers — the ones paying actual enterprise money — report the same “we’ll get back to you” template, followed by weeks of nothing.
– An India-based user: *”It’s been more than a year for us. We’ve resorted to using OpenRouter.”*
The recommended workaround, according to multiple commenters, is to find Anthropic staff on Twitter and publicly beg. That is not a customer service strategy. That is a hostage situation with better branding.
The uncomfortable part
Yesterday we wrote about how Claude’s pricing model is a game of “unlimited until we notice.” Today’s story is the other shoe dropping. When the pricing goes wrong — and it clearly *does* go wrong, unpredictably, at scale — there is nobody on the other end of the line to fix it.
An AI company that can’t implement a support inbox is asking developers to trust it with production workloads. Make that make sense.
And look, I get it. Anthropic is growing fast. Support is hard. Humans cost money. But if you’re charging people $200 a month for a “Max” plan while your frontline support is a chatbot and an unread inbox, that’s not a growing-pains problem. That’s a choice.
Here’s the part I keep coming back to. The only thing that got Nick’s money back was publishing the story. Not the support system. Not the escalation process. Not the three polite follow-ups over the course of a month. Just pure, public embarrassment.
If that’s what it takes, then the support system isn’t broken. It doesn’t exist.
All opinions expressed on BluntAI are editorial opinions based on publicly available information and personal testing. We may earn affiliate commissions from links on this site.
Disclaimer: BluntAI may earn affiliate commissions from links in this article. This never influences our reviews. We buy and test everything ourselves. Our opinions are brutally our own.