A minimal, transport-agnostic, opt-in HTTP standard for declaring and settling the energy, environmental, and human-attribution components of a single Large Language Model inference request.
Published in the spirit of the early IETF RFCs that defined TCP, IP, HTTP, and TLS — small, composable, opt-in, and owned by no single party.
- Read the spec →
spec/aiisp-1.md - Read it rendered → https://97115104.github.io/aiisp-spec/
- Read the companion paper → https://humandatacollective.org (commentable)
- Discuss v0.1 → issue #1
- Propose a change → open a PR or file an amendment
.
├── spec/ # Normative specification documents
│ ├── README.md
│ └── aiisp-1.md # AIISP-1 — the inference transparency layer
├── paper/ # Companion paper (motivation, not normative)
│ ├── human-data-collective.tex # LaTeX source (canonical)
│ ├── human-data-collective.md # Markdown mirror
│ ├── human-data-collective.pdf # Typeset edition
│ ├── human-data-collective.bib # Bibliography
│ └── generate-latex.sh # pdflatex × 3 build script
├── docs/ # GitHub Pages site (index + spec renderer)
├── .github/ # Issue templates, PR template, Pages workflow
├── CONTRIBUTING.md # How to file an issue or open a PR
├── GOVERNANCE.md # How decisions about the standard are made
├── CHANGELOG.md # Revision history
├── CITATION.cff # Machine-readable citation metadata
├── CODE_OF_CONDUCT.md # Contributor Covenant v2.1
├── SECURITY.md # Private security disclosure
├── LICENSE # CC BY 4.0 — applies to spec & paper text
└── LICENSE-CODE # MIT — applies to scripts & reference code
This is a Draft (v0.1, April 2026) circulated for public comment. It is not, at the date of writing, a ratified standard of any standards-development organization. The author intends to submit this document to the IETF ART area, the Decentralized Identity Foundation, and the W3C AI Provenance community group, in that order.
The protocol family is called AIISP. This specification is AIISP-1, The Artificial Intelligence Inference Standards Protocol. Future specifications will be AIISP-2 (Attribution), AIISP-3 (Governance), and so on, following the IETF RFC numbering convention.
The repository hosts two things you can contribute to: the spec in
spec/aiisp-1.md and the companion paper in
paper/. Both follow the same PR-and-issue workflow against
different files.
| You want to… | Use this |
|---|---|
| Comment generally on AIISP-1 v0.1 | Reply on issue #1 |
| File a specific objection on a section | Comment template |
| Propose a new feature, header, or behaviour | Amendment template |
| Report a typo or broken citation | Errata template |
| Propose textual changes | Open a Pull Request |
| Comment on a specific paper passage | Select text at https://humandatacollective.org |
| Disclose a security issue privately | See SECURITY.md |
Full guidelines are in CONTRIBUTING.md. Decision
process and versioning are in GOVERNANCE.md.
cd paper && ./generate-latex.shRequires TeX Live or MacTeX (brew install --cask mactex-no-gui on macOS).
See paper/README.md for the full editorial workflow.
- Specification text in
spec/and paper text inpaper/are licensed Creative Commons Attribution 4.0 International so any party may implement, mirror, translate, or fork without permission. - Scripts and any reference implementation code are licensed under the MIT License.
@misc{aiisp1_2026,
author = {Harshberger, Austin},
title = {AIISP-1: The Artificial Intelligence Inference Standards Protocol},
year = {2026},
month = apr,
version = {Draft v0.1},
howpublished = {Open standard, GitHub repository},
url = {https://github.com/97115104/aiisp-spec},
note = {Public review thread: \url{https://github.com/97115104/aiisp-spec/issues/1}}
}Plain prose:
Harshberger, A. (2026). AIISP-1: The Artificial Intelligence Inference Standards Protocol (Draft v0.1). https://github.com/97115104/aiisp-spec
This project follows the Contributor Covenant v2.1. By participating you agree to abide by its terms.
Editor: Austin Harshberger · x@97115104.com