AI governance research substrate
AI governance substrate is infrastructure that governs AI behavior outside the model's own self-reporting or prompt compliance. Prompted LLC works in the software/logical substrate layer: runtime governance for AI harnesses that need to know when to act, when to ask for human judgment, and how to turn that judgment into safer autonomy over time.
The unified frame
Prompted LLC builds Ubiquity so AI systems can know when to act, when to ask, and how to turn human judgment into safer autonomy over time.
AI governance substrate is infrastructure that governs AI behavior outside the model's own self-reporting or prompt compliance. Prompted LLC works in the software/logical substrate layer: runtime governance for AI harnesses.
Context Grapple Gun is the portable governance lifecycle that proves Ubiquity's grammar inside Claude Code. It captures lessons from real work, routes them through human review, promotes them through scoped gates, and hydrates approved guidance back into future sessions — without confusing rendered context with constitutional source of truth.
CGG is complete without Ubiquity. When flat-file governance is no longer enough, Ubiquity extends the same primitives into deeper substrate infrastructure: semantic recall, graph topology, conformation-aware retrieval, expression gating, and compiled execution-boundary enforcement.
Why prompt compliance and post-hoc guardrails are not enough
Prompt compliance asks the model to behave. Output filtering catches the model after it has already produced text. Neither shapes what the system is permitted to do before action becomes real. For agentic systems acting across tools, sessions, and time, governance has to live in the runtime — the layer where intent, memory, authority, and consequence actually meet.
Four substrate paradigms
Different governance problems live in different substrate paradigms. Treating them as one category is what produces vague policy that no layer can enforce.
How Ubiquity occupies the software/logical layer
Ubiquity is the runtime governance engine for AI harnesses. It treats purpose, trust, risk, memory, and human judgment as primitives — not as principles to apply after the fact. The lifecycle is Signals → Warrants → Rules: provisional lessons surface, hold under review, and only become durable when evidence stabilizes.
The substrate does not remove people. It moves human judgment to the seams where it is decisive, and carries the answer forward so the same decision is not asked again on the next session.
How CGG proves the grammar in Claude Code
Context Grapple Gun is the developer-facing proof of the substrate grammar. It runs locally over flat files: capture lessons from real work, route them through human review, promote them through scoped gates, hydrate approved guidance back into future sessions. The lifecycle is complete without Ubiquity. When flat files run out, the same primitives extend into the deeper substrate.
What this is not
AI governance substrate is not a compliance dashboard. It is not RLHF. It is not a prompt-filter layer. It is not a hardware attestation product. It is not a workflow automation platform. Those surfaces address adjacent problems and compose alongside the substrate — they do not replace it.
Demand ladder
This is
- A software/logical governance substrate for AI behavior.
- Runtime governance for AI harnesses.
- Infrastructure for governed autonomy.
- A category map for AI-mediated operating reality.
This is not
- Prompt engineering.
- RLHF.
- A compliance dashboard.
- A model wrapper.
- A hardware attestation product.
- An MLOps platform.
- An AI safety blog.
Frequently asked
- What is an AI governance substrate?
- Infrastructure that governs AI behavior outside the model's own self-reporting or prompt compliance. The substrate shapes what AI systems can do at the moment of action, rather than relying on post-hoc filtering or policy documents the agent never sees.
- How is this different from AI guardrails or Constitutional AI?
- Guardrails and constitutional methods operate in or around the model. A governance substrate operates underneath the harness — in the runtime layer that determines what may become action — so trust, risk, memory, and human judgment have weight before output is produced.
- Why split substrates into four paradigms?
- Different governance problems live in different layers. Hardware substrates govern compute and attestation. Software/logical substrates govern runtime state, memory, and gates. Institutional substrates govern human/AI authority. Measurement substrates govern evidence. Conflating them produces vague policy that no layer can enforce.
- Where does Prompted LLC fit?
- Prompted LLC works in the software/logical substrate lane. Ubiquity is the runtime governance engine; Context Grapple Gun is the portable governance lifecycle that proves the grammar inside Claude Code. The other paradigms are composed alongside, not replaced.
- Is decentralization required for an AI governance substrate?
- No. Decentralization is a deployment choice, not a definition. A substrate can be centrally operated, federated, or fully local. What makes it a substrate is that governance lives below the agent, not above it.