A universal governance standard for human oversight of artificial intelligence.
"AI may analyze, recommend and simulate — but a human must authorize."
The Judge Protocol is a proposed global governance framework that automatically monitors, approves and manages AI execution across multi-level environments — enabling humans to retain control over all consequential AI decisions.
The framework proposes:
- An International Artificial Intelligence Agency (IAIA) — a UN specialized agency modeled on the IAEA, with national implementation bodies (NAIA) in every member state
- HACK — Hardware ACK — a hardware enforcement mechanism embedded in GPU firmware, implementing a chip-level gate that locks AI compute capacity until a verified human authorization token is received
- A four-tier Decision Gate — a structured human authorization mechanism for all consequential AI actions
- PACK — Physical ACK — a two-part human authorization for the physical domain, covering robotics and autonomous weapons
- ACP — Action Classification Protocol — a lightweight descriptor attached to every compute instruction, providing context for the HACK chip before execution
Built on the same architectural principle as TCP/IP: just as no packet is considered delivered without an explicit ACK, no consequential AI action is considered authorized without an explicit human acknowledgement.
| Document | Description | DOI |
|---|---|---|
| Executive Summary v0.91 | Start here — 9-page overview of the framework | 10.5281/zenodo.20401666 |
| The Judge Protocol v0.93 | Full framework — governance structure, decision gate, hardware enforcement, domain rules, funding model | 10.5281/zenodo.20425549 |
| Technical Architecture HLD v0.63 | Hardware and software architecture — HACK chip, COE, ACP, three-layer stack, phased implementation | 10.5281/zenodo.20460111 |
No AI action is considered authorized until an explicit human acknowledgement is received.
| Tier | Classification | Authorization |
|---|---|---|
| Tier 1 — Hard Stop | Existential, permanently irreversible | HARD STOP — no override |
| Tier 2 — High | Irreversible, large-scale | Multi-party human authorization |
| Tier 3 — Significant | Material impact, partially reversible | Single designated human authority |
| Tier 4 — Routine | Standard, reversible, low impact | AI executes autonomously — logged |
Working draft — open for review and contribution. The framework is designed to evolve.
Published by the Judge Protocol Foundation, Stockholm, Sweden.
Creative Commons Attribution 4.0 International (CC BY 4.0)
Hardware architecture designs released under CERN Open Hardware Licence Version 2 — Strongly Reciprocal (CERN-OHL-S v2)
© 2026 Alexander Hofmann. You are free to share and adapt this material for any purpose, provided appropriate credit is given.