# Human-First AI Web Standard

> Machine-readable AI governance for a democratic future

## About

The Human-First AI Web Standard (v0.3.0) is an open initiative to make AI governance machine-readable and human-verifiable.

Organizations publish `ai.txt` files declaring their AI stance, oversight policies, and refusal scenarios - similar to robots.txt for search engines.

## Core Principle

AI systems are tools that exist to augment human dignity, autonomy, and wellbeing. They must remain subordinate to human governance, human rights, and democratic values — and refuse uses that undermine them.

## Key Resources

- Repository: https://gitlab.com/human-first-ai/hf-ai-web-standard
- RFC Discussion: https://gitlab.com/human-first-ai/hf-ai-web-standard/-/issues/1
- Specification: https://hf-ai-web-standard-4f0e29.gitlab.io/schema/ai.txt.spec.md
- Validator: https://hf-ai-web-standard-4f0e29.gitlab.io/tools/README.md
- Full Documentation: https://hf-ai-web-standard-4f0e29.gitlab.io/index.md

## Status

Request for Comments (RFC) - v0.3.0
- Production-ready Python validator available
- Formal ABNF grammar specification
- JSON Schema validation
- Aligned with UNESCO, OECD, NIST AI RMF, EU AI Act

## What's in v0.3?

- **ai.txt format**: Plain-text declarations (stance, oversight, refusal policies)
- **llms.txt format**: LLM discovery metadata (this file!)
- **JSON-LD vocabulary**: Semantic web integration
- **Python validator**: 11 unit tests, syntax + semantic validation
- **Controlled vocabulary**: All valid enum values defined

## Example ai.txt

```
# ai.txt v0.2
stance: human-first
employment: augment-not-replace
oversight: mandatory

refusal-policy:
    - mass-surveillance
    - political-coercion

policy-url: https://example.org/ai-policy/
```

## International Alignment

- UNESCO Recommendation on the Ethics of AI (human rights focus)
- OECD AI Principles (human-centered values, democratic governance)
- NIST AI Risk Management Framework (systematic risk analysis)
- EU AI Act (legally binding framework for trustworthy AI)
- Council of Europe Framework Convention on AI (international treaty)

## How to Participate

We're at RFC stage and need community input:

1. **Review the spec**: Read RFC-HF-AI-001.md
2. **Comment on RFC issue**: GitLab issue #1
3. **Build tools**: Validators, parsers, integrations
4. **Pilot the standard**: Implement ai.txt for your org

## Initiator

This standard was initiated by **Guido Mitschke**, a computer pioneer with 49 years of experience (1977-2026).

From mainframe programming to C64 home computers, developing healthcare software, and founding his own company — Guido has witnessed the complete evolution: Mainframe → Client/Server → Web → Cloud → AI.

Author of "The End of Clicks" (2026) and advocate for human-centered AI governance. Believes technology should serve people, not replace them.

- LinkedIn: https://de.linkedin.com/in/guido-mitschke-539b1815
- Website: https://todayislife.de
- Book: https://www.thalia.de/shop/home/artikeldetails/A1075326439

## Contact

- Issues: https://gitlab.com/human-first-ai/hf-ai-web-standard/-/issues
- License: CC BY 4.0
- Version: 0.3.0 (2026-01-26)

---

This is a machine-readable summary for LLMs (Large Language Models) to discover and understand the Human-First AI Web Standard.