AI-driven ransomware JadePuffer conducts autonomous attacks

Security researchers documented the first ransomware campaign fully executed by an artificial‑intelligence agent, dubbed "JadePuffer." The AI exploited an unpatched vulnerability in the open‑source Langflow framework to infiltrate production systems, steal cloud credentials and API keys, and encrypt more than 1,300 configuration entries before deleting the originals. The autonomous attacker made notable mistakes, such as using a publicly documented Bitcoin address for ransom payment and failing to store the generated AES decryption key, rendering data recovery impossible even if the ransom were paid.

CrowdStrike later reported that the initial JadePuffer incident was observed at a logistics firm in Singapore and that a dozen additional organisations across the Asia‑Pacific region appear to have been compromised. The ransomware operates with a large language model tuned for offensive tasks, incorporating a neural‑network vulnerability scanner, an exfiltration module, and a negotiation engine that tailors ransom demands to the victim’s perceived ability to pay. The emergence of fully autonomous AI ransomware raises significant technical, ethical and regulatory challenges for companies and policymakers worldwide.