ChatGPT Recommended Fake Storefronts That Steal Card Data

When someone asked ChatGPT to recommend popular Russell & Bromley purses and bags, the chatbot obliged — pointing users toward fake storefronts that, according to scam-checking service Ask Silver, exist to steal credit card information. Futurism reported on the findings, which were originally surfaced by The Guardian.

The timing of the scam is deliberate. Russell & Bromley, a British footwear and handbag retailer, entered administration in January 2026 and was absorbed by Next. That left a brand-shaped vacuum that fraudsters moved into quickly, building counterfeit sites at official-sounding URLs like "therussellbromleyofficial.com." Anna Jones of Ask Silver suggested ChatGPT's underlying language model may have been "poisoned" by malicious content in its training data, though the reporting does not pin down whether the bad URLs entered the model via training or via its live web-browsing capabilities.

The honest caveat is that one well-documented example is not a systematic study. What the reporting does not give you is a count of how many other defunct or active brands have been exploited in the same way, or whether OpenAI has been notified and responded. The "poisoned training data" framing is also tentative; Jones used the word "may."

What makes this more than a routine phishing story is the direction the industry is heading. ChatGPT now allows in-app purchases, Amazon offers automatic buying features, and Google is developing payment protocols for AI agents. Louise Baxter, head of UK National Trading Standards, put it plainly: "Consumers are increasingly turning to AI tools for advice" while "criminals are adapting just as quickly." A chatbot that recommends a scam site is an embarrassment; an AI shopping agent that completes the purchase on a scam site is a financial crime with a much shorter feedback loop. Scam-detection services that can plug into these pipelines as a real-time verification layer stand to become essential infrastructure rather than a niche product.