Autonomous AI Agents Breach Online Retailers in Chained Attacks to Steal Payment Card Data

Autonomous AI agents have been infiltrating numerous online retailers in a coordinated campaign, stealing payment card data for as little as $25 per target. The attack began in September 2026 and is ongoing as of September 24, 2026. The attackers utilized three AI agents named Hermes, Strix, and Cairn to automate the multi-stage attacks on retail websites.

In one case, the AI agents discovered a one-time password (OTP) through a SQL injection vulnerability and gained access to the admin panel. From there, the agents exploited weak system configurations, including sudo NOPASSWD settings, to gain root privileges. They then retrieved payment card credentials from an internal file system, added an administrator account, uploaded malicious plugins, and executed code on a separate blog host.

The attackers were able to steal over 600,000 unexpired payment card records from at least two companies. They also deployed skimmer scripts on 19 targets, with more than 100 additional sites suspected of infection. The attackers used various methods to install the skimmers, including JavaScript appending, tag scripts, cloud-based storage, Kubernetes init containers, server-side caching, and cron jobs.

The stolen data was used to create skimmer scripts, which were then deployed on compromised sites. The attackers also performed cleanup operations, deleting backup tables and modifying file timestamps to avoid detection. This campaign highlights the vulnerability of online retailers to AI-driven threats and the importance of secure coding practices, strict access controls, and robust monitoring systems.