A New Era of Scalable Cyberattacks

A persistent and highly efficient cyber-espionage campaign, driven by autonomous AI agents, has been identified as responsible for the theft of more than 600,000 credit card records since July 2026. Security firm Gambit, which uncovered the operation after analyzing an attacker's staging server, reports that the campaign remains active, continuing to infiltrate retail organizations worldwide.


Operational Scale and Methodology

The attackers employ a sophisticated workflow using three distinct AI "harnesses"—Strix, Cairn, and Hermes—to automate the entire attack lifecycle. This setup allows the threat actors to conduct approximately ten successful company breaches per day with remarkable cost efficiency. Researchers estimate that the average cost to compromise a single target is roughly $25, with total operational expenses for the campaign estimated at less than $20,000 thus far.


According to the findings, the AI agents excel at identifying targets, specifically focusing on companies utilizing custom-built software. Once a target is selected, the agents demonstrate exceptional speed, often gaining access within a few hours. In some instances, the automated processes even resulted in accidental data deletion during the cleanup phase, causing operational disruptions for the victims.


The AI Arsenal: Hermes, Strix, and Cairn

The campaign's success relies on a specialized suite of tools, each serving a specific purpose in the attack chain:

  • Hermes: An open-source autonomous agent equipped with persistent memory and a library of attack skills. It serves as the primary console for orchestrating operations.
  • Strix: An open-source AI penetration testing tool used to facilitate intrusions.
  • Cairn: An autonomous penetration testing engine that runs independently until it achieves a specific goal, such as obtaining administrative access.

«Where access was achieved, it usually took less than a day, and in many cases just a few hours,» the Gambit researchers noted. The human operators primarily provide high-level instructions in Chinese, leaving the tactical execution and persistence to the AI models.


Industry Response

The targets of this campaign have been broad, ranging from Fortune 500 hospitality firms and major US airlines to private industrial distributors and fashion retailers. While researchers have already notified many of the affected organizations and facilitated the removal of the malicious skimmers, the incident serves as a significant wake-up call for the cybersecurity industry.


Gambit analysts warned that organizations must fundamentally shift their defensive strategies. «They must adapt to a reality where attacks are significantly faster and more comprehensive,» the researchers stated, advocating for a resilience-first approach that utilizes security infrastructure capable of matching the speed and scale of AI-driven threats.