Anthropic cancelled the wrong org, kept ~$3,900 of what we paid, and support says the decision is "final"

Detailed Analysis

A Reddit post detailing a protracted billing dispute between a business customer and Anthropic has surfaced as a cautionary account of how the company's support infrastructure handles errors at the enterprise level. According to the poster, a 30+ seat Claude Team/Enterprise customer, an internal Anthropic bug created a duplicate organization that was billed alongside their legitimate account earlier in the year. After the customer disputed the duplicate charge via chargeback, Anthropic's response was to cancel the wrong organization — the real, active one with roughly 315 of 365 subscription days remaining — rather than the duplicate. The customer repurchased immediately to avoid disrupting their team, then continued adding seats to the reinstated organization over subsequent weeks, accumulating roughly CA$9,900 in invoices. Anthropic refunded only the base subscription invoice (~CA$5,000) and left approximately CA$4,900 in seat add-ons unrefunded, despite those seats belonging to the same cancelled org.

The core of the dispute is an accounting sleight-of-hand: Anthropic appears to be crediting the refund from the duplicate org's chargeback against the real org's losses as well, effectively using one refund to offset two distinct financial harms. The customer says they provided a detailed invoice-by-invoice spreadsheet with live formulas and repeatedly offered to accept Anthropic's own numbers if the company would simply prorate the cancelled seats rather than only the base plan. Despite this, support responses reportedly took weeks, were frequently generated by an AI agent rather than a human, refused to issue a case number after three requests, and ultimately declared the decision "final." Notably, the original duplicate organization that prompted the whole episode remains uncancelled four months later, undercutting Anthropic's claim that the matter was properly resolved.

This account matters because it exposes a gap between Anthropic's rapid growth in enterprise and business API/subscription revenue and the maturity of its billing and customer-support operations. As Anthropic has scaled Claude for Work, Claude Enterprise, and API-based business accounts to compete with OpenAI's ChatGPT Enterprise and Microsoft Copilot, it has taken on obligations more typical of established SaaS vendors — accurate invoicing, seat management, dispute resolution, and escalation paths for high-value accounts. A billing error compounded by an incorrect cancellation, followed by an unwillingness to reconcile a documented discrepancy, suggests support processes not yet built for the complexity of multi-seat, multi-invoice business accounts, especially when errors originate on the company's own infrastructure rather than customer misuse.

The situation also illustrates a broader tension increasingly visible in the AI industry: companies racing to convert consumer-grade chat products into enterprise revenue streams while support, billing, and trust-and-safety systems lag behind. The customer's dilemma — needing to choose between recovering disputed funds via another chargeback or risking account suspension and losing access to a tool they depend on daily — highlights a power asymmetry common when AI vendors control both the product and the sole channel for dispute resolution, with limited regulatory or contractual recourse for business customers. As more organizations embed Claude and similar models into daily operations, incidents like this are likely to draw attention to whether AI vendors' operational and support maturity is keeping pace with their commercial ambitions, and whether "unjust enrichment" from billing errors becomes a recurring friction point industry-wide rather than an isolated incident.