Detailed Analysis
A Korean-language Reddit post translated into English details a compounding customer service failure experienced by an Anthropic subscriber who was allegedly hacked through the platform's "Max 20 gifting feature" and charged $220 without authorization. The user's subsequent month-long attempt to obtain a refund was entirely mediated by Fin, a third-party AI chatbot deployed by Anthropic as its primary customer support interface. Fin reportedly acknowledged the refund request and claimed it would forward the matter to the relevant team, but no follow-up communication was ever sent. When the user returned a month later seeking a status update, Fin had no record of the prior conversation — a critical failure in continuity that left the user with no recourse and no documentation of their case within Anthropic's own system.
The most operationally damning element of the account is the circular logic embedded in Anthropic's support architecture. When Fin directed the user to navigate to "Account > Get Help" to submit a refund request, that pathway opened Fin itself — the same chatbot that had already failed to resolve the issue. The user's attempt to escalate via a direct email to Anthropic's user safety address resulted in an automated or cursory response that reiterated the same "Account > Get Help" instruction before closing the case entirely. This design effectively creates a closed loop in which users with legitimate billing grievances are cycled repeatedly back to the same unresponsive interface, with no visible path to a human representative or a meaningful resolution mechanism.
The situation escalated further when the user discovered that Anthropic's platform appeared to be attempting to process a subscription renewal charge — for a plan the user had already canceled — when they logged back in after a month away. Because Anthropic reportedly does not provide users with a self-service option to delete stored payment card information, the user was forced to cancel the physical card itself to prevent any unauthorized charges. These two issues — the attempted charge on a canceled subscription and the inability to remove payment credentials — represent substantive data privacy and consumer protection concerns that go beyond mere customer service friction. The user ultimately resolved the billing dispute through their credit card company rather than through Anthropic, which reflects a complete breakdown in the company's ability to handle fraud-adjacent cases internally.
These complaints situate Anthropic within a broader tension that is increasingly visible across the AI industry: companies scaling their consumer-facing products rapidly while relying on automated support systems that are structurally incapable of handling the edge cases — fraud, unauthorized charges, account compromise — that require human judgment and institutional accountability. Anthropic's use of Fin, a product built by Intercom, as a frontline and apparently sole support mechanism stands in notable contrast to the company's public positioning around Claude as a highly capable, trustworthy AI assistant. The irony the user explicitly names — that Claude is marketed as sophisticated while Fin fails basic support continuity tasks — points to a gap between Anthropic's research and product reputation and its operational infrastructure. The BBB's F rating, referenced by the user, suggests this experience is not isolated and that Anthropic has accumulated a pattern of unresolved consumer complaints that regulatory and consumer protection entities have begun to reflect in formal assessments.
Read original article →