Customers of the Claude artificial-intelligence chatbot are being targeted by a fraud scheme in which scammers use compromised accounts to buy hundreds of pounds’ worth of gift cards, The Guardian reported.

David Duggan, a Claude user on the U.S. East Coast whose name was changed for the report, told the newspaper he discovered the scheme after his wife spotted two $200 charges on his credit card bill for gift cards tied to the chatbot, purchases he had not made. A third $200 attempt was blocked because it required confirmation, he said.

“My wife asked me: ‘Hey, did you make these $200 purchases?’ It was $400 in total. And then there was a third one, but that needed a confirmation and did not go through,” Duggan told The Guardian. He said he had originally signed up for a $20-a-month subscription to Claude after being impressed with its ability to answer medical questions and help organize family life.

According to The Guardian, when Duggan contacted Anthropic, the company behind Claude, his account was suspended, but automated replies failed to explain what had happened. After searching online, the paper said, he found other users describing similar experiences on Reddit. One user reported 10 unauthorized payments of £18 each while another said they had been charged €216 three times, and two more reported being unexpectedly billed €225 apiece.

The genuine gift-card vouchers were subsequently sent to Duggan’s personal email, raising fears that his email account had also been compromised and could expose him to additional scams. After he updated his credit card details online, two further attempted payments were blocked, the newspaper said, and he is now working with his bank to recover his money.

The fraudulent transactions appear on bank statements as payments to Anthropic. One Reddit user shared an email reading “You’ve received a gift!” along with a link to redeem a Claude subscription, according to the report.

Anthropic told The Guardian it is putting new protections in place to prevent fraudulent gift-card purchases, and that when it identifies scam transactions it cancels the subscriptions and issues refunds. The company said there was no evidence that compromised card details had originated from Anthropic itself.