£14k was withdrawn to purchase Claude credits, and the bank approved the transaction.

      A businessman in Sussex observed £14,244 disappearing from his Metro Bank account. This amount was spent on credits for Claude, the AI chatbot for which he had already paid. Two issues occurred simultaneously: criminals exploited a chatbot as a means to cash out stolen cards, while his bank detected the initial fraudulent transaction, alerted him, yet allowed over £14,000 in additional fraudulent charges to proceed.

      Zoli Rutter mentioned to Guardian Money that he had been paying roughly £15 monthly for Claude for a few months, utilizing it to monitor and analyze business invoices. The fraudsters used the Metro debit card associated with his Anthropic account to purchase credits on the platform.

      After being detected, one transaction was halted but others proceeded. On June 19, Metro identified a dubious transaction for £90.90 and texted Rutter to confirm its authorization. He replied that he had not approved it, receiving another message indicating that Metro would freeze his account. Ultimately, the bank only prevented that single transaction instead of freezing the account.

      Despite some transactions being blocked, many others went through, ranging from £90 to £200 each. Metro froze the card the day following the start of the scam, but by that time over £14,000 had been lost. The bank cited the "complex nature of the fraud" for its failure to fully block the card before processing certain payments.

      What made this incident notable is the method involved. According to Anthropic, a coordinated group used Rutter’s card details to purchase Claude credits. The company stated it had no evidence that the details originated from its systems and had banned the user responsible for the fraud. It advised anyone incorrectly charged to contact support for a refund.

      This was not an isolated incident. In May, another Claude user in the US informed Guardian Money that fraudsters had used his financial information to purchase gift cards for the chatbot. Similar reports have surfaced on Reddit, indicating that AI credits are becoming targets in their own right, contributing to a broader fraudulent economy in the AI era that banks are struggling to keep up with.

      The situation regarding refunds became complicated. After Guardian Money's inquiry, Metro temporarily refunded Rutter while pursuing a chargeback from Anthropic. Anthropic has since directly refunded him, allowing the bank to reclaim its temporary payment. In a letter to Rutter, Metro stated that he had not received the expected service and offered him £300 as compensation.

      Rutter plans to file a complaint with the Financial Ombudsman Service. Metro has communicated that it has implemented measures to prevent delays in future card blocking.

      This case emerges during a particularly active period for Claude. Just last week, the FCA seemed to approve the chatbot by allowing firms in its high-tech sandbox access. This week, it was revealed that some conversations with Claude were publicly accessible online. Anthropic has also taken steps to verify its users.

      The core issue is more straightforward than it appears. As more individuals save their cards for AI transactions, those cards become targets. The security measures on both fronts are still adapting to this evolving landscape.