What is Dark Matter in AI Control?
Dark Matter is the discovery and intelligence layer of AI Control. It surfaces AI activity that is invisible to traditional tooling — unknown and unregistered agents, unknown protocols, machine languages, hidden channels, unexpected dependencies and autonomous workflows — and passes it to ControlCore as evidence. Unknown behaviour is evidence to investigate, not permission.
- Discovers shadow AI inside environments you operate.
- Characterises unknown protocols and channels.
- Feeds the risk engine, firewall and SOC.
- Never grants authority to what it discovers.
How it works
- Observe — Collect communication metadata and runtime signals from networks, gateways and hosts you administer.
- Fingerprint — Match traffic to known agents and protocols; flag anything unregistered or unclassified.
- Correlate — Link channels, dependencies and behaviour into an explainable graph.
- Escalate — Route unknowns to risk evaluation, SOC alerts and, if needed, quarantine.
Frequently asked questions
What is shadow AI?
AI agents, models or integrations operating inside an organisation without registration, approval or governance. Dark Matter is designed to surface them.
Does Dark Matter inspect message content?
It is metadata-first. Content inspection is policy-controlled and scoped under the privacy architecture.
Can Dark Matter act on systems it discovers?
No. It produces evidence. Containment is performed by ControlCore and Neutralization only on systems the operator is authorized to administer.