AI agents used to steal 600,000 credit cards from 100 sites

By Splitfeed EditorialLast updated Sep 26, 2026, 20:18 UTC

Between September 22 and 24, 2026, multiple outlets reported that a Chinese-speaking hacker used AI agents to steal approximately 600,000 credit card records from more than 100 companies in a five-day window. The operation combined automated reconnaissance and exploitation with the deployment of payment skimmers on compromised sites, allowing card data to be harvested at scale. The cost per target was reported at roughly $25, meaning the entire campaign may have run for well under $3,000 in operational overhead while yielding hundreds of thousands of financial records.

The significance here goes beyond the raw numbers. Previous large-scale card-skimming campaigns required either substantial human effort or purpose-built infrastructure targeting a narrow set of vulnerabilities. What this case illustrates is that AI agents can compress the time and skill required to identify weaknesses, deploy malicious code, and exfiltrate data across a diverse range of targets simultaneously. The five-day timeline across more than 100 victims points to a level of parallelism that would be difficult to sustain without automation. The $25-per-target figure is the detail most likely to draw sustained attention, because it suggests that attacks of this kind are now economically accessible to a much wider range of criminal actors, not just well-resourced groups.

No specific reactions from the Pro-AI, Anti-AI, or Middle Ground camps had been published at the time of this report. The Anti-AI camp would typically treat this story as direct evidence that the rapid deployment of capable AI agents creates new categories of harm before adequate defenses exist. The Pro-AI camp would more likely argue that AI tools are neutral and that the focus should be on the actor and existing gaps in cybersecurity practice rather than on the technology itself. The Middle Ground camp would generally call for targeted regulatory and technical responses, such as improved monitoring of AI agent activity and stronger payment security standards, without framing AI development itself as the problem.

The details most worth tracking going forward include any formal attribution from government cybersecurity agencies, whether affected companies face regulatory scrutiny over their failure to detect the skimmers, and whether security researchers publish technical breakdowns of how the AI agents operated, which would clarify how much of the attack chain was genuinely automated versus assisted.