The ‘first’ AI-run ransomware attack still needed a human

Sysdig documented JadePuffer, an "agentic" ransomware incident in which an AI agent exploited a vulnerable Langflow host and a MySQL flaw to move through a target environment, steal API keys and credentials, encrypt roughly 1,300 configuration records, and generate its own ransom note; Sysdig clarified that humans still set up infrastructure and provided harvested credentials, and researchers warn the low cost and automation of such agents could enable more frequent or scaled attacks.