AI Agents Used to Support E-Commerce Attacks and Steal Credit Card Data

* Researchers from Gambit Security disclosed an attack campaign using open-source AI tools: Strix, Cairn, and Hermes.
* The campaign was active starting in July 2026.
* Between September 10 and 15, researchers observed 105 attack activities.
* At least 27 organizations were confirmed compromised to varying degrees.
* More than 600,000 unexpired payment card records were stolen from two companies.
* The attack involved using Strix for vulnerability identification, Cairn for testing/exploiting vulnerabilities, and Hermes for control/coordination.
* One observed path included SQL Injection leading to system access, web shell upload, privilege escalation, and data access on AWS.
* Web skimmers were deployed to steal payment card information from checkout pages on 19 websites and over 100 other linked websites.
* Administrators should monitor JavaScript files, checkout pages, AWS S3, and databases for changes.
* Attackers deleted stolen data after successful information theft.

The mechanism described illustrates a convergence where accessible, democratized tools—open-source AI agents—are being leveraged to operationalize complex cyberattacks against commercial targets. The pattern involves abstracting the steps of penetration testing and exploitation (vulnerability discovery via Strix, exploitation via Cairn) and adding an orchestration layer (Hermes), suggesting that the barrier to entry for sophisticated attack execution is decreasing through tool aggregation rather than requiring specialized skill sets alone. This shifts the focus from individual vulnerability management to systemic oversight of the tools themselves, as evidenced by the necessity for administrators to monitor data storage locations like AWS S3 and databases where exfiltration occurs. The pattern of utilizing web skimmers highlights that the final objective often involves direct, automated data theft from transaction points. The implication is a necessary shift in organizational defense priorities from perimeter security to internal data flow integrity and continuous environmental monitoring of code and infrastructure.
What mechanisms exist for auditing the provenance and operational use of these integrated AI agents within an organizational context? How does the reliance on ephemeral, open-source tools change the structure of accountability when attacks utilize novel combinations of existing capabilities? What are the long-term societal implications of payment data being exposed via automated means facilitated by publicly available software?