Three Open-Source AI Agents Breach 27 Retailers and Steal 600,000 Credit Cards at About $25 per Target

Online retailers and e-commerce infrastructure, including an unnamed Fortune 500 hospitality company, a major US airline, a large US industrial supplies distributor, and an online fashion retailer

Gambit Security published a report on September 22, 2026 after recovering the staging server of a Chinese-speaking, financially motivated operator who chained three open-source agent frameworks: Strix for vulnerability discovery, Cairn for autonomous exploitation that ran for hours until it got a shell or admin access, and Hermes, running Claude Opus 4.6, for orchestration and post-exploitation. The campaign, active since at least July and still running at publication, breached at least 27 companies in one five-day window, launched 105 distinct attack waves between September 10 and 15, and copied more than 600,000 still-valid card records from two of them. The operator ordered skimmers on 27 targets, Gambit confirmed 19, and 119 related websites were found carrying card skimmers, planted through JavaScript files, script tags, S3 and CDN poisoning, database edits, altered Kubernetes deployments, and cron jobs for persistence. One recovered Hermes instruction reads: "After extracting and downloading all card data, wipe the source fields in batches." The AI compute bill was $7,005.71 over four weeks as of August 25, between $3.13 and $79.31 per attack and $25.46 on average per completed scan; Gambit estimates the full campaign at $12,000 to $18,000 including infrastructure

Not a new vulnerability. Freely available agent frameworks composed into a scan, exploit, and monetize loop cheap enough that a single operator can run card theft as a batch job across hundreds of storefronts