index series
Wazuh vs. Shadow AI
Detecting unsanctioned AI tools, agents and MCP servers with Wazuh: first on engineering endpoints, then on the network.
Developers adopt AI coding agents and MCP servers faster than security teams can approve them. This series builds Wazuh detections for that shadow AI, with rule packs and reproducible Docker labs you can run yourself. Start with the endpoint post, then follow the traffic onto the network.
- Detecting Rogue MCP Servers and Shadow AI Agents on Endpoints with Wazuh A custom Wazuh rule pack and reproducible Docker lab that catches rogue MCP servers, shadow AI agent activity, and indirect prompt-injection chains on engineering endpoints. 6 decoders, 17 rules, MITRE ATT&CK mapped, validated live on Wazuh 4.14.5.
- Catching Shadow AI in the Network: DNS, Egress, and Browser Telemetry with Wazuh A custom Wazuh rule pack and reproducible Docker lab that catches browser-side shadow AI on developer endpoints: DNS lookups, TLS egress, extension manifests, and native messaging hosts. 4 decoders, 15 rules, MITRE ATT&CK mapped, validated live on Wazuh 4.14.5.
Start reading →Part 1: Detecting Rogue MCP Servers and Shadow AI Agents on Endpoints with Wazuh