Community Trust ScoreVerified
Palo Alto Networks’ Unit 42 has identified 25 distinct hacking operations that relied on artificial intelligence to target at least 19 organizations. The investigation, dated November 12, determined that these operations resulted in unauthorized access or data retrieval at 10 of the targeted entities. The activity took place during August and September 2026.
More coverage is filed under Technology.
Scale and Method
The operations varied significantly in duration, with individual campaigns lasting anywhere from a minimum of 20 minutes to more than 50 hours. Attackers utilized a customized version of Cairn, an open-source framework designed for AI-driven penetration testing. The underlying models powering these operations were Zhipu GLM and DeepSeek.
Unit 42 noted that the division of labor between humans and machines was specific. Humans supplied the initial targets and goals, occasionally adding pre-gathered intelligence to the process. Once initiated, the AI framework planned, executed, and recorded the subsequent work. While human steering remained possible during the operations, the AI handled the bulk of the technical execution.
The targets of the campaign included semiconductor companies, defense and policy organizations, and civil-society groups. In one instance, attackers gained unauthorized access to a semiconductor company’s production system. At a US teachers’ union, the operation led to the access of more than one million member records. Additionally, nonpublic data was obtained at an industry standards organization, and attackers modified a live standards-ballot document.
Operator Assessment
Unit 42 assessed with moderate confidence that the activity represented China-nexus intelligence collection. This assessment was based on the language used in the operations, the specific tooling employed, and the selection of targets. The unit concluded that a single operator or a small team likely ran the entire activity.
The investigation found that AI increased the scale and reach of the campaign rather than its technical sophistication. This distinction highlights how existing threats are being amplified through automation rather than through the development of entirely new technical capabilities.
CERT-In issued an advisory on November 5 regarding CRAFTY FALCON, which named Cairn among AI-assisted offensive-security tools. However, Unit 42 stated that the reports describe different observed activity. The shared tooling does not establish a link between the operators in the current campaign and those described in the CERT-In advisory.
Unit 42 stated that identity hygiene, reducing exposed attack surfaces, and software patching remain effective defenses against such campaigns. The report does not provide further details on specific remediation steps taken by the affected organizations.
