Automated Assaults on Global Retailers

A persistent and highly effective cyber espionage campaign, driven by autonomous AI agents, has been targeting retail organizations worldwide since July 2026. According to findings from security firm Gambit, the attackers have successfully compromised numerous companies, resulting in the theft of at least 600,000 payment records. The campaign remains active, continuing to breach new websites even as security teams work to mitigate the threat.


Gambit researchers gained insight into the operation after discovering and analyzing the attacker's staging server. The evidence revealed that, within just a five-day window in September, the AI agents executed over 100 attack waves, successfully compromising 27 organizations. Among those targeted are prominent US-based entities, including a Fortune 500 hospitality firm, a major airline, and a large-scale industrial distributor.


Weaponizing AI for Low-Cost Attacks

The threat actors, believed to be financially motivated groups based in China, have developed a highly efficient attack model. By leveraging three specialized AI "harnesses"—named Strix, Cairn, and Hermes—the attackers can automate almost the entire exploitation chain.


The economic efficiency of this operation is notable. Gambit reports that the attackers spent approximately $7,000 over four weeks, bringing the total cost per target to an average of roughly $25. Researchers noted the extreme speed of these incursions, stating:

«Where access was achieved, it usually took less than a day, and in many cases just a few hours. We also detected instructions in the attacker's playbook that could disrupt the operations of a company as a result of data deletion or cleanup procedures run by the agent - and this has indeed happened in some of the breaches.»

The AI Arsenal

The campaign utilizes a sophisticated suite of tools to manage and execute attacks:

  • Hermes: Serves as the central console for orchestration. It utilizes advanced AI models to execute human-provided instructions, which are brief commands in Chinese that guide the agent's next objectives after an initial breach.
  • Strix & Cairn: These function as an open-source pentesting tool and an autonomous engine, respectively. They are designed to identify targets, launch exploits, and achieve administrative access without human intervention.

A New Era of Cybersecurity Defense

Gambit researchers expressed significant concern regarding the scalability and persistence of this campaign, noting that the AI's ability to remain patient and creative exceeds the capabilities of most human operators. The team emphasized that the cybersecurity landscape has fundamentally shifted.

Organizations are now urged to adopt a "resilience-first" strategy. To defend against these high-speed, automated threats, companies must deploy security infrastructure capable of matching the velocity and comprehensiveness of AI-driven attacks. While many impacted victims have been notified and malicious skimmers removed, the ongoing nature of the threat highlights the urgent need for more adaptive defense mechanisms.