1 Star: Premium subscriber treated as a non-customer
I'm a paying Claude Pro Max 5x subscriber at the highest consumer tier. Here is what happened today, with timestamps so the events speak for themselves.
At 12:54 I sent Anthropic support a detailed email about sustained service quality issues with Claude Code. The email was around 1,500 words and included two evidence attachments totalling 18 pages: a technical incident report on a regression in the latest Claude Code VS Code extension (versions 2.1.141 and 2.1.142), and a candid self-report Claude itself wrote acknowledging a separate session that destroyed a documentation migration on my project. The email cited six prior support tickets, three of which Anthropic's AI agent ('Fin') had explicitly escalated to a human with the message 'we'll email you as soon as an agent has responded.' No human ever replied to any of those three. My email explicitly asked, after six AI-only conversations and three undelivered human-escalation promises, for a response from a human.
At 12:57, three minutes later, I received a reply declining the refund. It is not physically possible for a human to have read either attachment, the email body, the six prior conversation transcripts, or the GitHub issue cluster cited as evidence in three minutes. The reply is signed at the bottom 'This answer was composed by Fin AI Agent, Anthropic's AI Agent.' The Fin signature confirms what the timing already proved.
I replied within the hour, calmly, addressing each point Fin made and giving Anthropic until end of business the following day to put a human on the case before I initiated a chargeback and a complaint to the Irish consumer protection authority. Their response, again within minutes and again from Fin, was to close the ticket entirely: 'This conversation has been closed and is no longer monitored. If you need further assistance, please submit a new request.'
So to be clear about what a premium subscription with Anthropic buys at the moment:
- A technical regression openly documented across 30+ separate user reports on Anthropic's own public GitHub repository, present and unresolved for at least 2.5 weeks. Other users describe it as 'breaks about 50% of my chats or more' and 'I'm not receiving most of my messages.' Anthropic's automated triage bot has been closing these reports as duplicates of an unrelated bug it claims is 'fixed' in the current version. It isn't.
- Documented support failure across multiple contacts. Fin AI handles every interaction. When Fin says a human will follow up, no human does. When the customer escalates with detailed evidence, Fin replies in three minutes and closes the ticket.
- A cancellation flow that doesn't actually let you cancel. My billing settings show only a 'downgrade to Pro' option — there is no cancel button. Under EU consumer law, cancellation must be at least as accessible as subscription. Right now it isn't.
- Extra-usage charges burning through the monthly limit on retries and re-prompts caused by the broken service. I've spent roughly 60% more than the headline subscription rate this cycle and the bulk of it has gone to redoing work the service failed to deliver the first time.
The product itself, when it works, is genuinely the best AI assistant on the market. I have defended it in conversations with other developers. But the gap between the quality of the model and the quality of the support is now wide enough that I cannot in good conscience recommend it to anyone. The 1.4/5 Trustpilot rating across hundreds of reviews exists because this is the standard experience, not the exception.
If you are considering subscribing to Claude Pro or Pro Max, please read the existing reviews carefully before committing your card details. The product is excellent. The support is non-existent. The cancellation path may not work when you need it. And if you have a problem with billing, technical service, or anything else, you will be talking to Fin AI, and Fin AI will not help you.
