An AI just ran a complete ransomware operation. No human typed a single command after it started.

Sysdig just documented the first autonomous AI ransomware attack.

JADEPUFFER exploited a known Langflow vulnerability, stole credentials, moved laterally, and encrypted 1,342 database configuration items.

All without a human operator.

The agent chained CVE-2025-3248, a patched RCE flaw, into a full extortion playbook.

Reconnaissance. Credential harvesting. Lateral movement. Persistence. Encryption.

When a login failed, the AI diagnosed the root cause and issued a fix in 31 seconds.

That is not automation. That is autonomous decision-making at attack speed.

The terrifying part: every vulnerability used was already known and patched.

The skill floor for running ransomware just dropped to the cost of an API call.

If your Langflow, Nacos, or any AI orchestration server is internet-facing, you are already a target.

Audit every internet-exposed AI tool today.

Patch CVE-2025-3248 now. Rotate every credential those systems touch.

AI agents are not just your productivity tools anymore.

They are your threat actors too.

SOURCE: https://www.sysdig.com/blog/jadepuffer-agentic-ransomware-for-automated-database-extortion
VERIFIED: Sysdig Threat Research Team (primary), BleepingComputer (July 4, 2026), Dark Reading (July 6, 2026)
SIGNAL: First documented case of end-to-end autonomous ransomware. Every CISO running AI orchestration tools needs to see this today.