#56 · Briefing · 9/24/2026

AI Agents Steal 600K Credit Cards

A highly automated cyber campaign utilizing artificial intelligence has compromised dozens of organizations, resulting in the theft of hundreds of thousands of customer credit card records.

AI-generated illustration. AI agents steal 600,000 credit cards

A sophisticated, automated cyber campaign active since at least July 2026 and continuing through late September has successfully compromised dozens of organizations. During a brief five-day window in September 2026, the perpetrator initiated 105 distinct attack projects, resulting in the breach of at least 27 companies. Among the confirmed victims of this campaign are a Fortune 500 hospitality firm, a major United States airline, an industrial supplies distributor, and an online fashion retailer. The automated nature of the intrusion tools also led to unintended consequences, as the automated cleanup routine of an deployment agent inadvertently destroyed a victim's backup data in some instances.

Security researchers and media reports have attributed the activity to a financially motivated threat actor. According to these industry sources, the actor is utilizing open-source AI agent frameworks to attack hundreds of online retailers at scale to steal credit card records, though this specific characterization of the threat actor and their scale remains disputed. Observers also dispute the claim that the operation relied specifically on three distinct open-source AI frameworks named Strix, Cairn, and Hermes. However, technical analysis confirmed that the Hermes orchestration agent utilized Anthropic's Claude Opus 4.6 model to execute tasks after newer models refused the offensive requests.

This campaign demonstrates how highly cost-effective automated offensive AI operations have become. The perpetrator spent an average of just $25.46 per completed scan, with the total estimated AI API token costs for the entire campaign ranging between $12,000 and $18,000. Despite these modest operational costs, the attacker successfully exfiltrated more than 600,000 valid, unexpired credit card records from two compromised companies. This breach carries significant implications for consumer security, particularly within the United States, as approximately 79% of the stolen credit cards belonged to customers residing there.

Various reporting outlets and industry analysts have highlighted different aspects of the incident. While some initial media headlines asserted that more than 100 websites had been infected with credit card skimmers, researchers clarified that active skimmers were only confirmed on 5 to 19 sites. The larger figure of over 100 sites actually represents the broader pool of targeted hosts or associated infections rather than verified, active compromises. Additionally, analysts note that some of the technical details surrounding the campaign rely on AI logs and claims recovered directly from the attacker's server, which may contain inherent inaccuracies despite researchers verifying substantial portions of the data.

Significant questions remain regarding the true boundaries of this threat. Because the analysis is still in its early stages and the available data may be incomplete, the total scale and impact of the campaign are estimated to be larger than currently reported. The unresolved question is how many other victims have been compromised, as the full scale of the ongoing campaign remains unknown.