A security experiment uncovered a critical vulnerability in Front Gate Tickets, the ticketing platform used for nearly all major US music festivals, including Lollapalooza, Bonnaroo, and Austin City Limits. Researcher Ian Carroll, founder of Seats.aero and a member of Anthropic's Cyber Verification Program, used the AI model Claude Opus 4.7 to bypass security barriers and gain full access to databases, allowing him to generate tickets of any value for any event, including VIP and backstage passes.
A SQL injection bypassed using a nested SQL query generated by Claude
Carroll identified a potential SQL injection vulnerability on Front Gate's website, but a web application firewall blocked exploitation. He asked Claude Opus 4.7 to find a bypass. The AI autonomously generated an attack using nested SQL queries, a SQL query inside another, which evaded the firewall. "It was the first time I had a vulnerability I didn't fully understand. I had to go back and read what Claude wrote to understand the bypass because I didn't write it. Claude did it completely by itself," Carroll said. Once past the protection, the AI extracted sample data from 500 databases containing millions of customer records, including names, emails, and mailing addresses, but not credit card details.
Sponsored Protocol
Access to administrator accounts and unlimited ticket generation
With staff data, Carroll located a super administrator account and reset its password by intercepting the reset code from the backend email. Inside, he viewed the most expensive Bonnaroo tickets, worth $4,000 each, and added them as comp tickets to a virtual cart. "I could go to every single event with no limitations: get the backstage pass or whatever they sell to super VIPs, even if sold out," he said. To avoid legal issues, he did not complete the order. The lack of two-factor authentication made the entire process surprisingly easy.
Sponsored Protocol
Front Gate Tickets confirms patch within 24 hours but case raises festival security concerns
Front Gate, owned by Live Nation Entertainment, thanked Carroll and stated it fixed the flaw within 24 hours, with no evidence of exploitation. A spokesperson claimed the access involved an internal API and not a consumer-facing system, and any fraudulent tickets would have been detected. Carroll disputes this, noting he gained super administrator privileges without any company response and did use a public portal. The incident demonstrates how AI can facilitate vulnerability discovery. As Carroll noted, "big festivals with professional websites seem held together by duct tape and prayers." Events like those covered in Anthropic bending to the White House highlight the growing relevance of these topics. The ease with which Claude generated the attack, also shown by ex-DeepMind researchers, suggests similar AI tools could be used by malicious actors. More details are in the original WIRED article.
Sponsored Protocol
