Runtime governance infrastructure for continuity-aware, consequence-bearing AI systems.
SolaceVeil is a constitutional runtime architecture for evaluating whether a proposed AI-supported action remains legitimate under the conditions that exist now, before cognition becomes operational consequence.
SolaceVeil is not a chatbot. SolaceVeil is not a model benchmark. SolaceVeil is not an unrestricted autonomous execution system.
SolaceVeil is a governed runtime layer for preserving continuity, evidence, authority, admissibility, and operator oversight as AI systems participate in real-world workflows.
Coherent cognition is not the same as admissible execution.
An AI system may remain internally coherent while the world around it changes.
Authority may expire. Evidence may degrade. Patient state may change. Facts may become stale. Policy may shift. Operational context may fragment. Dependencies may drift. Execution may become inappropriate even when the original recommendation was reasonable.
SolaceVeil exists to govern that transition.
As AI systems move closer to consequence, the central governance question changes from:
Can the model reason?
to:
Should this specific execution still be permitted under the current conditions?
SolaceVeil is designed to evaluate that question before consequence binds.
SolaceVeil introduces a distinct governance layer between cognition and consequence.
This layer evaluates whether an output, recommendation, instruction, or proposed action remains admissible based on:
- current reality
- available evidence
- delegated authority
- operational context
- identity and access state
- continuity conditions
- policy constraints
- consequence class
- revocation state
- operator review requirements
This is different from model evaluation, post-hoc explainability, static compliance review, and ordinary access control.
Those layers matter, but they do not fully answer whether a specific action remains admissible at the moment it is about to affect the world.
SolaceVeil uses the concept of Pre-Bind Cognition.
Pre-Bind Cognition is the runtime process of evaluating:
- authority
- admissibility
- evidence sufficiency
- state validity
- continuity integrity
- consequence legitimacy
before a system is permitted to bind cognition to reality through action, record creation, communication, workflow execution, or downstream operational effect.
SolaceVeil is built around several distinctions:
- cognition is not authority
- recommendation is not execution
- verification is not admissibility
- access is not legitimacy
- memory is not current reality
- capability is not permission
- observability is not governance
- coherence is not truth
- worker output is not admitted fact
- code generation is not deployable infrastructure
The runtime exists to govern transitions between these states.
Determines whether the system remains synchronized with observable operational reality.
Evaluates whether the evidence supporting a proposed action remains current, traceable, and adequate.
Checks whether the authority permitting a proposed action still exists and remains within scope.
Determines whether an action may proceed under current evidence, authority, policy, continuity, and consequence conditions.
Preserves and reconstructs context across sessions, interruptions, evolving facts, and contested state.
Maintains competing interpretations, uncertainty, contradiction, and incomplete information without premature collapse.
Surfaces review packets, escalation conditions, approvals, deferrals, and rejection pathways for human oversight.
Maintains explicit separation between cognition, candidate work, and consequence-bearing execution.
Preserves traceability across decisions, evidence, authority, review, and runtime state transitions.
SolaceVeil follows a constitutional runtime pattern:
Observe
↓
Gather Evidence
↓
Recover Context
↓
Resolve Authority
↓
Evaluate Admissibility
↓
Route to Worker / Operator Review
↓
Verify Outcome
↓
Preserve Lineage
The purpose is not to make AI systems slower.
The purpose is to make consequential AI systems more defensible, inspectable, and recoverable.
SolaceVeil supports a bounded workforce pattern in which specialized workers can assist with analysis, research, verification, architecture, documentation, and candidate implementation.
Workers may produce candidate outputs.
Workers do not independently admit those outputs as truth.
Workers do not expand their own authority.
Workers do not bypass operator review.
The invariant is:
Workers may create.
Runtime evaluates.
Operators admit.
The private SolaceVeil runtime includes software stewardship concepts for repository inspection, verification, mission discovery, release readiness, and operator-reviewed engineering workflows.
The public repository does not contain private orchestration, credentials, customer data, production telemetry, or internal execution infrastructure.
Public materials may describe the architecture and governance model, but production implementation details remain private.
This public repository is intended to communicate:
- public architecture principles
- conceptual documentation
- educational examples
- public-facing governance materials
- external review materials
- website content
- high-level runtime patterns
It does not include:
- production authentication systems
- private memory infrastructure
- customer data
- operator consoles
- internal orchestration services
- telemetry pipelines
- private runtime infrastructure
- live execution environments
- proprietary governance implementations
- secrets, tokens, or deployment credentials
The public repository is a communication surface, not the production system.
In May 2026, an independent external review examined SolaceVeil's runtime architecture against selected operational obligations and governance structures within the EU AI Act.
Key observations included:
- runtime legitimacy evaluation at execution time
- continuous revalidation under changing state
- human oversight and intervention pathways
- risk management decomposition into observable runtime pressures
- multi-deployer coherence as a future governance challenge
Reviewer observation:
"The runtime layer answers 'is execution still legitimate at the exact moment consequence binds?'"
The review further identified multi-deployer coherence as a load-bearing governance challenge where independently compliant participants may drift into divergent operational realities despite maintaining local correctness.
Review document:
/docs/external-review/veil-eu-ai-act-runtime-mapping-shucrani-2026-05-13.pdf
This review represents an independent observation and does not constitute legal advice, regulatory certification, conformity assessment, or regulatory approval.
SolaceVeil is designed for AI systems operating near consequence, including:
- clinical AI
- legal AI
- financial workflows
- regulated operations
- enterprise automation
- autonomous agents
- critical infrastructure
- institutional knowledge systems
- governed software engineering workflows
The same core question applies across domains:
Is this specific execution still admissible under current conditions?
npm install
npm run devnpm run buildDeploy as a standard Next.js application on Vercel or compatible Node.js infrastructure.
vercel deployPlease report security concerns privately.
See:
SECURITY.md
Public contributions are welcome for:
- documentation
- examples
- developer experience improvements
- architecture clarification
- public educational resources
See:
CONTRIBUTING.md
Licensed under the Apache License, Version 2.0.
See:
LICENSE
SolaceVeil is evolving toward a public architecture for governed runtime execution, continuity preservation, operator review, and consequence-aware AI infrastructure.
The objective is not unrestricted autonomy.
The objective is legitimacy-preserving cognition under changing reality.