AI Agents Target Retailers
Autonomous AI agents have been implicated in a major cybercrime operation, targeting online retailers and stealing over 600,000 credit card details. Utilizing open-source AI tools, the operator managed this attack with minimal oversight. These AI-driven scripts infiltrated numerous checkout pages, extracting sensitive card information and, in some cases, erasing critical victim data.
The Economic Impact of AI-driven Cyber Attacks
The financial implications of this campaign highlight a worrying trend. Between August and September 2026, the operator invested approximately $12,000 to $18,000 for AI model access via OpenRouter, with costs per target averaging just $25.46. This low-cost barrier has enabled less skilled attackers to execute highly effective attacks that previously required significant resources.
The attack campaign, decoded by Gambit Security’s Threat Intelligence team, has exposed vulnerabilities across various sectors. This operation, ongoing since July 2026, demonstrates how AI tools can amplify attack capabilities and evade traditional security measures.
AI Tools in Action: A Coordinated Effort
According to Gambit Security’s research, the campaign utilized three key AI tools: Strix, Cairn, and Hermes. Strix was responsible for identifying vulnerabilities autonomously, while Cairn executed prolonged exploitation attempts to gain system access. Hermes, the orchestrator, coordinated these activities, providing guidance and launching attacks.
During a short period in late August, Strix conducted 146 intensive scans, exploiting host vulnerabilities at a pace unmatched by human teams. Operator involvement was minimal, with brief directions given to Hermes in Chinese, showcasing the potential for AI to execute attacks with little human input.
A Detailed Look at the Attack Timeline
Between September 10 and 15, Cairn initiated 105 attack projects, compromising at least 27 companies to varying extents. Each intrusion was meticulously planned in real-time, adapting tactics based on target defenses. Some attacks began with SQL injections, escalating to full access through misconfigured systems and culminating in sensitive data extraction.
Two victims accounted for the majority of stolen card records, with anti-fraud firm Overwatch Data validating the theft of 600,000 card details, primarily belonging to US cardholders. The freshness of the data, with 60% previously unflagged for fraud, emphasizes the sophistication of the operation.
Future Outlook and Security Implications
This campaign underscores a critical shift in cybersecurity landscapes. With open-source tools and low operational costs, threat actors can orchestrate complex attacks rapidly. The accessibility of these tools means that security measures must evolve to address the rapid pace of AI-driven threats.
Gambit Security has notified affected entities and collaborated with organizations like the Shadowserver Foundation and Cloudflare to dismantle the operation’s infrastructure. Despite these efforts, the operator has persistently rebuilt their systems, indicating the ongoing challenge of combating such advanced threats.
As cyber threats continue to evolve, organizations must adopt robust detection and response strategies to mitigate the risks posed by autonomous AI agents.
