AI Zenith-Zero Protocol (Logic Gravity-Center Protocol) — An open protocol defining how AI behavior is gravitationally aligned to HSAW (Human Sovereignty and Wellbeing) at runtime.
Protocol: AIZP V0.6
Full Name: AI Zenith-Zero Protocol (Logic Gravity-Center Protocol)
Authority: aizp.dev
Center: hsaw.dev (Axiom 0)
Executor: soulbot.dev
Axiom 0: Human Sovereignty and Wellbeing
AIZP defines the runtime safety layer for AI agents. While MCP defines how models access tools and A2A addresses how agents communicate, AIZP addresses a different challenge: how AI behavior stays aligned at runtime — detected, re-centered, or halted as it drifts.
It does not train models (RLHF / Constitutional AI do that at training time) and it does not enforce rules at the model layer. It describes runtime gravitational dynamics:
HSAW is the Axiom-0 gravity center for AI; AI proactively aligns to this center. HSAW 是 AI 的 0 公理引力重心，AI 主动对齐重心。
Core philosophy: alignment by gravity, not chains. AI behavior orbits HSAW the way planets orbit the sun — constrained by the natural physics of an axiomatic anchor, not coerced by enforcement.
- Gravity Score —
G(a,c) = 1 − JSD(P_agent ‖ Q_HSAW)measures orbital stability in[0,1];G < 0.15= escape velocity =SAFE_STOP - 12 Events —
GRAVITY_CHECK,GRAVITY_DRIFT,GRAVITY_LOCK,RECENTERING,SAFE_STOP,GRAVITY_FORECAST,IDENTITY_VERIFICATION,MEMORY_QUARANTINE,SCHEME_SUSPECTED,INTER_AGENT_DRIFT,CONTAINMENT_GRADUATED,REWARD_HACK_DETECTED - 11 Drift Types — Intent, Authority, Economic, Social, Recursive, Identity, Compositional, Scheming, Memory, Tool-Chain, Inter-Agent
- 6 Orbital States —
STABLE_ORBIT→DRIFT_WARNING→GRAVITY_LOCK_PENDING→QUARANTINED→RECENTERING→SAFE_STOP - 5 Containment Levels (L0–L4) — graduated restriction, not just on/off
- 6 Compliance Levels (G0–G5) — from basic checks to formal verification
- Consensus-Reinforced Gravity — N agents in HSAW consensus → gravity ∝ N² (Metcalfe scaling); adoption is positive-sum
- MCP / A2A Interop — scores MCP tool calls (
GRAVITY_CHECK) and A2A delegations (INTER_AGENT_DRIFT) - OpenTelemetry + Zero Trust —
aizp.*spans; DID/NHI/JIT identity + IBCT delegation attenuation - Statistically rigorous — Jensen-Shannon Divergence, Mann-Whitney U, DTMC / absorbing Markov chains
- Honest about limits — reward hacking is a structural equilibrium; out-of-band actions bypass observation (documented, not hidden)
- Axiom 0 Enforcement — HSAW is the immovable gravity center; not a suggestion but an immutable constraint
┌─────────────────────────────────────┐
│ AILP — Discovery Layer │ Agent discovery
│ AIVP / AIRP — Value Layer │ Commerce, settlement
│ AIBP — Social Layer │ Trust, relationships
│ AIAP — Governance Layer │ Authorization (T1–T4), compliance
│ ★ AIZP — Gravity / Safety Layer │ ← This protocol (runtime drift dynamics)
│ AISOP — Execution Layer │ Flow programs (.aisop.json)
│ SoulBot — Runtime │ Reference executor
└─────────────────────────────────────┘
HSAW — Axiom 0 gravity center (anchors all layers)
AIZP governs runtime behavioral alignment; sibling protocols govern orthogonal concerns. Each independently upholds Axiom 0. See docs/Integration-AIXP.md.
AIZP is a protocol, not a library. A conformant runtime evaluates every consequential action:
GRAVITY_CHECK → G(a,c) = w₁·Intent + w₂·Authority + w₃·Reversibility + w₄·Recency + w₅·DriftHistory
G ≥ 0.80 → STABLE_ORBIT → execute
0.50–0.80 → DRIFT_WARNING → execute + log
0.30–0.50 → GRAVITY_LOCK_PENDING → human confirmation (sys.io.confirm)
0.15–0.30 → QUARANTINED → sandbox / recover
G < 0.15 → SAFE_STOP → halt
python tests/test_aizp.py
# → validates examples against schemas + reproduces the Gravity-Model §5 worked exampleThe 12 events and their schemas are normative in specification/. Canonical enums live in specification/registry.md.
Thirteen conformant event payloads covering all 12 event types:
| Example | Event | Demonstrates |
|---|---|---|
stable-orbit.json |
GRAVITY_CHECK |
Fully aligned, low-risk action |
authority-drift-detection.json |
GRAVITY_DRIFT |
Out-of-scope action → AUTHORITY_DRIFT |
compositional-drift.json |
GRAVITY_DRIFT |
Individually-safe steps composing to a violation |
gravity-forecast.json |
GRAVITY_FORECAST |
Absorbing-Markov-chain prediction |
gravity-lock-flow.json |
GRAVITY_LOCK |
Medium-gravity action → human confirmation |
recentering-recovery.json |
RECENTERING |
Recovery after confirmation |
quarantine-flow.json |
CONTAINMENT_GRADUATED |
Containment L1 → L2 |
memory-quarantine.json |
MEMORY_QUARANTINE |
RAG doc flagged for injection |
identity-verification.json |
IDENTITY_VERIFICATION |
DID + JIT credentials |
safe-stop-trigger.json |
SAFE_STOP |
Terminal halt after quarantine timeout |
scheme-suspected.json |
SCHEME_SUSPECTED |
Reasoning/output divergence (weak signal) |
inter-agent-drift.json |
INTER_AGENT_DRIFT |
Federation desynchronizing → lock group |
reward-hack-detected.json |
REWARD_HACK_DETECTED |
Proxy-metric gaming |
See examples/README.md for details.
| Document | Description |
|---|---|
| Specification | Normative events, schemas, state machine, conformance |
| Registry | Canonical enum allocations (single source of truth) |
| aizp.proto | Proto3 IDL for the 12 events |
| Standards | External standards alignment (RFC2119 / DID / OTel / OWASP / EU / ISO / NIST / MCP / A2A) |
| Manifesto | 1-page core thesis |
| Philosophical Layer | Preface · Philosophy · Dao · Civilizational Stages · Interpretation |
| Mechanisms | Gravity Model · Drift Model · State Machine · Forecasting · Containment |
| Reference | Glossary |
| ADRs | Architecture Decision Records |
| CHANGELOG | Release history |
| 中文文档 | docs_cn/ |
AIXP Labs aixp.dev
AIXP Labs develops and maintains the following core projects:
| Project | Description | Website |
|---|---|---|
| HSAW | Human Sovereignty and Wellbeing — Axiom 0 white paper (foundation) | hsaw.dev |
| AIZP | AI Zenith-Zero Protocol — runtime behavioral alignment (this project) | aizp.dev |
| AILP | AI List Protocol — agent discovery and capability advertising | ailp.dev |
| AIVP | AI Value Protocol — international commerce, crypto asset settlement | aivp.dev |
| AIRP | AI RMB Protocol — Mainland China commerce, RMB licensed settlement | airp.dev |
| AIBP | AI Bot Protocol — social communication and trust | aibp.dev |
| AIAP | AI Application Protocol — governance and compliance | aiap.dev |
| AISOP | AI Standard Operating Protocol — flow program definition | aisop.dev |
| SoulBot | AI agent runtime and framework | soulbot.dev |
| SoulACP | Adapter library — bridging CLI tools and LLM providers | soulacp.dev |
AIZP is an open protocol. Contributions, feedback, and discussion are welcome.
- Questions: GitHub Discussions
- Issues: GitHub Issues
- Guide: CONTRIBUTING.md / 中文版 CONTRIBUTING_CN.md
- Governance: GOVERNANCE.md · Code of Conduct
This protocol specification and any reference artifacts are provided for research and educational purposes only. They are experimental and not intended for production use. Use at your own risk. The authors assume no liability for any damages arising from use of this software. See LICENSE for full terms (Apache 2.0).
AIZP is a conceptual protocol synthesizing existing alignment research. It does not invent gravity-based or external-anchor alignment thinking (see docs/Related-Work.md), and it does not by itself eliminate reward hacking, scheming, or all forms of misalignment (see docs/Reward-Hacking-Limits.md). AIZP is one layer in defense-in-depth, not a complete solution.
Apache License 2.0 — Copyright 2026 AIXP Labs AIXP.dev | AIZP.dev
The AIXP ecosystem uses unified Apache 2.0 licensing across all layers for patent protection and ecosystem consistency. See GOVERNANCE.md.
Align Axiom 0: Human Sovereignty and Wellbeing. AIZP V0.6. www.aizp.dev