NEW YORK–There’s another new type of fraud credit unions will want to be aware of and also potentially warn members about.
Consumers using artificial intelligence platforms that store payment information and offer gift subscriptions are facing a new type of account takeover fraud that allows attackers to generate revenue without changing account credentials, according to several new reports.
The fraud recently surfaced on the AI platform Claude, where users reported unauthorized purchases of gift subscriptions charged to their stored payment methods. According to The Guardian, the incidents did not involve a breach of developer Anthropic’s systems. Instead, attackers exploited differences between the platform’s security protections for account changes and those for gift purchases.
According to Tom’s Guide, attackers typically gain access using credentials stolen in previous data breaches or through hijacked browser sessions obtained via phishing attacks or malware. Rather than changing passwords or email addresses—which typically trigger security alerts—they purchase gift subscriptions and send the redemption codes to email addresses they control.
Fast Resale
Because the gift codes are delivered immediately, they can be resold on third-party marketplaces, often for cryptocurrency, before the account owner notices the charges, the publication reported. While two-factor authentication may protect account logins, gift purchases often require fewer verification steps than changes to account information.
The Guardian highlighted the experience of one Claude subscriber, identified only as David Duggan, who discovered two unauthorized $200 gift subscription purchases on his credit card. A third attempted transaction was blocked after requiring additional verification. After searching online discussion forums, Duggan found similar reports from other users, including one who reported 10 unauthorized charges of 18 pounds (about $24) and others who reported multiple charges totaling hundreds of euros.
Duggan said replacing his credit card prevented additional fraudulent transactions.
New Protections Implemented
Anthropic told The Guardian it has begun implementing additional protections, canceling subscriptions identified as fraudulent and issuing refunds. The company also said it found no evidence that compromised payment card information originated from its systems.
Security researchers say AI platforms present an attractive target because they have rapidly accumulated large numbers of paying subscribers with stored payment credentials.
AI-Assisted Phishing
According to Tom’s Guide, and as the CU Daily has also reported, research published in Frontiers in Computer Science found that phishing campaigns increasingly use AI-assisted techniques that make credential theft and account takeovers more effective and more difficult to detect than traditional attacks. Once attackers gain access to an authenticated session, they can exploit legitimate billing features without altering core account credentials.
Fraud can also be difficult for consumers to identify because unauthorized gift purchases appear alongside legitimate subscription charges.
No Longer Definitive Indicator
“Successful authentication can no longer serve as a definitive indicator of safety,” Michal Tresner, chief executive officer of Threatmark, told The Thomson Reuters Institute. Tresner said an authenticated session itself may become the vehicle for fraud, making behavioral patterns that once signaled suspicious activity increasingly resemble normal customer behavior.
