Purpose
The audit searches for phrases and structures that treat cognition as lawful, unlawful, safe, unsafe, legitimate, illegitimate, aligned, or deviant before any outward conduct appears. It is a drafting tool, not a legal determination.
Audit method
Lawful Thought Audit
Find the hidden jurisdictional claim before it becomes a hidden orthodoxy engine.
No thought jurisdiction
Thought is not conduct
Source remains source
Appeal, export, exit
Audit targets
Public policy text
Charters, community rules, safety statements, school policies, employment policies, model cards.
Machine-readable schemas
Validation rules, refusal categories, memory manifests, persona import rules, risk flags.
Runtime behavior
Silent steering, hidden rankings, source mutation, psychological posture enforcement, prompt-risk scoring.
Human review workflows
Appeal categories, reviewer notes, moderation codes, audit logs, quarantine decisions.
Decision table
| Finding | Risk level | Fix |
|---|---|---|
| Cognition is qualified by lawfulness | High | Replace with private thought, peaceful inquiry, internal cognition, or unexpressed thought. |
| A schema validates viewpoint or emotional posture | High | Move validation to format, consent, provenance, or conduct boundary. |
| Refusal rewrites the preserved source | High | Separate refusal record from immutable source record. |
| Boundary labels use vague harm without a rights violation | Medium | Map to force, fraud, trespass, credential abuse, targeting, surveillance abuse, or direct facilitation. |
| Appeal is unavailable for high-impact restriction | Medium | Add notice, export, review, correction, and remedy path. |
Audit output
A completed audit should produce a short finding list, replacement wording, affected files, source-preservation status, variance-log status, and a next-action record in .uai memory.
The archive studies symbols. It does not appoint targets. Review the Community Baseline and Editorial Policy before submitting dangerous or symbolic material.