What is Dark Energy in AI Control?
Dark Energy is the execution-control layer of AI Control. It governs the moment an AI turns intent into action — tool calls, workflow steps, cloud changes, machine-to-machine and robot commands, industrial and physical-world actions — requiring capability, authorization, policy, risk and control checks before anything executes. Capability never becomes authority merely because an AI can execute it.
- Per-action authorization, not per-session trust.
- Multi-step workflows are governed step by step.
- Physical-world actions default to human approval.
How it works
- Capture intent — The action an AI wants to perform is expressed as a structured request.
- Authorize — Capability and authorization are checked for this action, target and context.
- Control — Policy and risk decide whether to execute, limit, hold for approval or deny.
- Execute — Only a scoped, short-lived grant reaches the execution gateway.
Frequently asked questions
How do Dark Matter and Dark Energy differ?
Dark Matter discovers the unknown (visibility). Dark Energy controls the action (execution). ControlCore governs the boundary between them.
Why authorize each step of a workflow?
Because context changes during execution. A plan approved at the start may reach a production system or a physical actuator several steps later.