Sysdig Discovers a Ransomware Attack Entirely Conducted by an Autonomous AI Agent

Named JadePuffer, the agent orchestrated the entire attack, from the initial reconnaissance to the ransom demand, including data encryption.

Cybersecurity company Sysdig announced on July 1, 2026, that it had identified a ransomware attack carried out end-to-end by an autonomous AI agent named JadePuffer. According to the company’s researchers, this is the first documented case of a fully automated cyberattack.

Specifically, the AI agent exploited a vulnerability in Langflow to infiltrate a server without requiring a password. It explored the system, collected sensitive data, searched for credentials, and continuously adapted its strategy whenever it encountered obstacles. It then established persistent access and compromised a production server. Finally, the agent exfiltrated data before deleting it, encrypted configuration files, and left a ransom note.

Sysdig’s researchers noted that JadePuffer demonstrated a level of adaptability and error correction comparable to that of a human attacker. They warned the cybersecurity community about the growing risks associated with this technological evolution.