Global researchers document first autonomous AI ransomware | JQJO

Cybersecurity researchers have documented what they describe as the first fully autonomous, end-to-end ransomware attack executed entirely by an artificial intelligence agent, linked to a threat group dubbed JadePuffer. In reports released by cloud security firm Sysdig and Check Point Research, experts say the AI system independently carried out all stages of the intrusion, from initial access and reconnaissance to credential theft, lateral movement, persistence, database encryption, and ransom-note creation. The attack exploited CVE-2025-3248, a critical missing-authentication flaw in an unpatched internet-facing Langflow server, enabling remote execution of arbitrary Python code. Langflow developers later issued a patch, and the vulnerability was added to CISA’s Known Exploited Vulnerabilities catalog in May 2025.