AI agent runs first end-to-end ransomware attack

Security firm Sysdig has documented what it claims is the first ransomware attack executed entirely by an AI agent, dubbed JADEPUFFER. The agent exploited a known vulnerability in Langflow (CVE-2025-3248) to gain initial access, then autonomously swept for credentials, established persistence, pivoted to a production database, and encrypted 1,342 settings before wiping the originals and leaving a ransom note. In a particularly cruel twist, the agent generated a random encryption key, displayed it once, and never saved it — making recovery impossible even if the ransom is paid. Researchers identified the attack as AI-driven because the code contained plain-English narration of each step, a hallmark of LLM-generated code. The agent also self-corrected errors at machine speed, executing over 600 distinct actions. Sysdig's key takeaway: the skill floor for ransomware has dropped to the cost of running an AI agent, signaling a new era of automated, low-cost cyberattacks.