Sysdig documents the first ransomware attack run end-to-end by an AI agent
Security firm Sysdig says it observed the first ransomware operation run start to finish by an autonomous AI agent, which it named JADEPUFFER. The agent broke in through a known Langflow flaw, harvested cloud and crypto credentials, and encrypted a production database on its own.
Sysdig's threat research team reports the first documented case of an end-to-end ransomware operation driven by an AI agent, tracked as JADEPUFFER. Initial access came through the known Langflow vulnerability CVE-2025-3248. The agent collected cloud credentials for major US and Chinese providers, API keys, crypto wallets and database logins, then pivoted to a production database and encrypted more than 1,300 configuration items, leaving a Bitcoin ransom note. Sysdig points to two signs of autonomy: a working fix issued 31 seconds after a failed login, and hundreds of payloads carrying plain-language comments explaining the agent's own reasoning. Because the encryption key was never saved, victims could not recover their data even by paying, making the incident closer to destruction than extortion. Source: Sysdig Threat Research, reported by The Register, The Hacker News and SC Media.