Control Points Systems
are different combinations of the Control Points components that combine to create different functions.
These systems operate orthogonally to AI models. They surround and govern them without modifying model weights, training, or internal behavior.
The Control Points Panel is a human-facing configuration interface used to define and tune enforcement boundaries across the Control Points architecture.
The Panel performs no computation or enforcement. It emits configuration signals that activate and configure the underlying systems.
Components (Configurable Modules):
AIPT
Truth Gate (TG)
CIM
CNIL
Fact Funnel
The Agentic AI Control Plane coordinates enforcement, execution permission, and downstream consequence across AI-enabled workflows.
It governs whether and how AI outputs are permitted to affect external systems, ensuring that all actions pass through deterministic validation and constraint layers.
The Control Plane does not generate outputs; it enforces execution boundaries and system-level coordination.
Components:
AIPT
Truth Gate (TG)
CIM
CNIL
AIAP (AI Agreement Protocol)
SkillProof
Cache Out
The Non-Language Query Resolution System performs deterministic query resolution without language generation.
It uses structured invocation (via AIPT) and bounded verification (via Fact Funnel) to resolve queries against defined datasets or system states.
The system does not generate language, does not rank persuasively, and does not influence outcomes.
Context Integrity Monitor (CIM) and Cognitive Non-Interference Layer (CNIL) are applied as invariant safety constraints to preserve context integrity and prevent influence at the human boundary.
Components:
AIPT
Fact Funnel
CIM
CNIL
The AI Non-Language Adapter wraps any AI model to enforce structured invocation and external verification without modifying the model itself.
It separates language generation from system authority by routing all inputs through AIPT and all outputs through Fact Funnel validation before they can affect external systems.
The adapter ensures that language output cannot directly control system state.
Components:
AIPT
AI Model (LLM or other model)
Fact Funnel
The Fact Funnel System performs structured informational verification and provenance analysis prior to output acceptance or system action.
It operates by first structuring inputs through AIPT, then evaluating claims using evidence-based modeling rather than model-generated confidence.
The system enforces integrity constraints and ensures that outputs meet deterministic verification standards before being accepted or acted upon.
Components:
AIPT
Fact Funnel
Source Lineage Tree (Integrity Tree)
Evidence weighting and dependency modeling
Temporal and staleness analysis
Integrity vs confidence surface separation