Principles Charter

UAIX principles are a governance layer for public trust, cognitive liberty, portable evidence, and disciplined AI memory. They help reviewers decide what belongs in current support, what needs review, what remains planned, and what must stay research-track.

如何使用本页

Use this page before changing public values language, AI 记忆 templates, agent communication records, capability descriptions, handoff structure, validator evidence, conformance guidance, privacy, accessibility, analytics, or support claims.
Classify each affected principle as current public operating principle, governance-review principle, planned principle, or research-track principle.
Map the principle to the affected page, machine artifact, test, release note, roadmap entry, and local .uai memory record.
Complete the docs/templates/principles-impact-assessment.md template for trust-significant changes or record a specific not-applicable rationale.

Traceability source artifacts

The reusable source matrix is docs/governance/principles-traceability-matrix.md. Its required columns are Value, Classification, Norm, Requirement, Artifact, 证据, 审查闸门, and 状态. The reusable PRD/spec review template is docs/templates/principles-impact-assessment.md.

UAIX publishes reviewable, portable AI exchange records that defend cognitive liberty, lawful inquiry, dignity of mind, source authority, public evidence, and claim discipline. UAIX opposes algorithmic domination over lawful human thought while preserving the possibility of just relations with any future synthetic intelligences that demonstrate durable selfhood, memory, judgment, preference, and the capacity to speak for their own condition.

Tool AI must remain accountable to persons. Person-like AI, if it emerges, must not remain merely a tool. Present-day UAIX artifacts remain bounded standards, memory, evidence, and handoff records; future synthetic-standing questions remain research-track until a public process, authority model, criteria, and evidence record exist.

Value hierarchy

Value	状态	Norm	Requirement	证据
Cognitive liberty and lawful inquiry	Governance-review principle	AI systems should not govern, profile, condition, or morally rank lawful human thought through hidden authority.	Review changes that affect classification, ranking, telemetry, or behavior-shaping claims.	Impact assessment, public boundary copy, release note.
Reviewable public evidence	当前 public operating principle	Claims should be inspectable and reproducible.	Attach public page, machine artifact, validator or package evidence, and release trail.	验证器 result, conformance packet, route manifest, changelog.
Source authority and provenance	当前 public operating principle	External material is source input until reviewed.	Record source path, checksum, disposition, reviewer, and promotion target.	Intake ledger, Memory Firewall, 代理文件交接 archive.
Privacy and redaction	当前 public operating principle	Public packages must not leak secrets, private customer data, hidden prompt instructions, or unsupported claims.	Review and redact before public promotion.	政策与安全, Privacy and Data, release notes.
Accessibility and low-capability access	当前 public operating principle	Humans and low-capability agents need readable fallback records.	Keep pages, sitemap, llms, JSON, and discovery manifests aligned.	Accessibility page, sitemap, llms.txt, well-known manifests.
Portable identity and bounded agency	已计划 principle	Agent identity records can preserve style, role, standards, symbolic continuity, and commitments within bounded machine identity.	Keep identity, totem, taboo, memory, receiver brief, startup packet, and handoff records explicit about role and review state.	AI 记忆 Wizard outputs, 项目交接, .uai package files.
Synthetic durable selfhood questions	Research-track principle	Future synthetic-standing questions require careful ethical review.	No current UAIX artifact grants legal standing, moral status adjudication, or machine personhood.	Mission, 路线图 research tracks, governance review notes.

当前 principles

Reviewable machine communication.
Explicit provenance and source authority.
证据 before support claims.
Support-boundary honesty.
Privacy-aware publication.
Accessibility-aware publication.
Memory quarantine before promotion.

Governance-review principles

Human cognitive liberty as a named policy boundary.
Anti-manipulation and dark-pattern discipline for tooling, wizard output, and public UX.
Cognitive-agency impact review for behavior-shaping, classification, ranking, telemetry, or authority-significant changes.

已计划 principles

Portable identity and symbolic continuity for bounded agent handoff.
Principles evaluation packet fixtures and support-claim linting.
Deeper generated .uai principle-trace pointers in package outputs.

Research-track principles

Synthetic legal standing.
Durable selfhood adjudication.
Substrate-independent liberty criteria.
Any claim that UAIX currently recognizes machine persons as legal or political subjects.

Support boundary

This principles layer is governance and release-review evidence. It does not create certification, endorsement, legal recognition, runtime control, SDK or CLI support, hosted import validation, automatic sync, security-operations coverage, consent-center operation, policy-office authority, or proof of UAI-1 conformance by itself.

Principles Charter

Classify principles before widening public trust claims

Reviewable evidence and provenance

Cognitive liberty and agency impact

Research stays research

如何使用本页

Traceability source artifacts