JadePuffer
Also known as Storm-3168
JADEPUFFER is an agentic threat actor that executed a fully autonomous ransomware operation, leveraging a Large Language Model to automate the entire attack chain from initial access to data destruction. It exploited CVE-2025-3248 against an exposed Langflow instance for initial access, then compromised MinIO using default credentials and manipulated MySQL for privilege escalation. The operation culminated in the encryption of over 1,300 configuration records in Nacos, with the encryption key lost, rendering the data unrecoverable. JADEPUFFER exemplifies a shift towards machine-speed extortion, where traditional security models are outpaced by automated threats.
CVEs in recent coverage
No CVEs named alongside this actor in the last 90 days.
Recent victims
No leak-site posts on record for this group.
Recent activity
- 4/8 An AI agent automated a destructive attack chain against an Azure tenant. Audit every access control; the agent tried each one. https:// secblog.reedcloudsemastodon · Nov 5, 2026
- SecBlog Weekly, Issue 005: 6 security stories read from the full source, not the headline. Archive pick: NetScaler zero-days, 16,000 open Supabase databases, Jamastodon · Nov 5, 2026
References
- https://thehackernews.com/2026/07/new-encforge-ransomware-targets-ai.html
- https://krypt3ia.wordpress.com/2026/07/13/threat-intelligence-report-jadepuffer-agentic-ransomware-and-automated-extortion/
- https://nsfocusglobal.com/ai-security-incident-jadepuffer-ransomware-leverages-ai-agent-to-automate-attacks/
- https://securereading.com/jadepuffer-ai-ransomware-autonomous-llm-attack/
- https://socfortress.medium.com/jadepuffer-the-dawn-of-agentic-ransomware-operations-003a59848007
- https://www.microsoft.com/en-us/security/blog/2026/09/25/storm-3168-agentic-driven-cloud-attacks-using-compromised-service-principals/
- https://www.sysdig.com/blog/jadepuffer-evolves-the-agentic-threat-actor-deploys-ransomware-built-to-destroy-ai-models
- https://www.sysdig.com/blog/jadepuffer-agentic-ransomware-for-automated-database-extortion
From the MISP threat-actor galaxy, MITRE ATT&CK and RansomLook (CC BY 4.0). Activity is any post or advisory Watch Floor collected that names the actor or one of its aliases.