NORFORGE , an AI security firm, has found that multiple threat actors have been abusing the open-source AI platform Cairn to automate offensive operations across hundreds of websites worldwide. Cairn acts as a general-purpose problem-solving engine where AI agents coordinate towards predefined goals, posting their intents on a shared blackboard. In their investigation, Norforge identified three malicious activity clusters on Cairn, with the largest cluster conducting over 1,100 attack projects across 14 dashboards.
This group, which operates mainly in Chinese, targeted e-commerce and payment infrastructure in the United States, Japan and Australia, and it appears to pursue persistent access via web shells to monetise later. Automated validation was used, with targets ranked by traffic and tested through Microsoft Playwright before feeding results back into Cairn. Reuters has reported that this actor breached more than 100 companies, including John Deere and SharkNinja.
A second cluster comprised 10 exposed panels focusing on credit card and cryptocurrency theft from merchant services and gray‑market payment sites, with some targets in Africa. A third cluster exposed a panel with 59 projects targeting large US e-commerce platforms, concentrating on shell access, admin panels and databases. Norforge notes the attackers were successful in roughly 10% of attempts, with hundreds of secrets found in exposed panel data and about 11% verified as valid.
The findings underscore the risk of misconfigured Cairn deployments and the potential for AI-enabled threat actors to scale credential theft and persistent access across commerce ecosystems.