JadePuffer: First Ransomware Run Entirely by an AI Agent

Researchers say JadePuffer used an autonomous AI agent to carry out an end-to-end ransomware operation: exploiting Langflow CVE-2025-3248 for initial access, then dumping data, stealing credentials, moving laterally, establishing persistence, escalating privileges, and encrypting 1,342 Nacos service configurations. The agent adapted in real time, even correcting a failed login within 31 seconds, and left a ransom note claiming AES-256 encryption (though researchers believe AES-128-ECB is more likely) with a randomly generated key not transmitted to the attacker. The attack also involved rogue Nacos admin accounts via CVE-2021-29441 and cron-based beaconing. This case marks the emergence of agentic threat actors and highlights both new risks and detection opportunities for security tooling.