# Policy on Artificial Intelligence & Large Language Models (LLMs)

The **SAME Analog Modular Ecosystem** is built on mathematical rigor, executable
verification, and total human accountability. While AI tools and Large Language
Models (LLMs) are widely used in modern development workflows, their integration
into this project is governed by strict boundaries to maintain technical
integrity and clear ownership.

## 1. Code & Specification Generation

The use of LLMs as personal assistance tools (copilots, syntax helpers, or
drafting assistants) is **permitted**, subject to two non-negotiable conditions:

### A. Manual Human Submissions Only

**No AI Bots or Automated PRs:** Every pull request must be authored, reviewed,
and manually submitted by a human contributor. Automated PR bots, LLM agent
pipelines, or machine-generated submissions will be closed instantly.

### B. Absolute Human Responsibility

* **You Own Every Line:** The creator of a PR is 100% accountable for all text,
  math, equations, CAD graphics, and code contained within their submission.
* **Must Defend the Engineering:** You must be capable of answering granular
  technical, mathematical, and architectural questions about your submission
  during code review.
* **Zero-Tolerance Excuse:** Defending a bug, broken assertion check, or flawed
  physics model with *"the AI generated it"* is considered an admission that you
  do not understand your own PR. Such PRs will be rejected immediately.

## 2. Discussions & Communication

Human-to-human clarity is vital when debating precision metrology and hardware
specs.

* **No AI-Generated Comments or Reviews:** The use of LLMs to generate issue
  responses, PR reviews, technical debate, or general commentary is **strictly
  prohibited**. Discussion threads must reflect genuine human reasoning.
* **Translation Exception:** Non-native English contributors are permitted to
  use machine translation to write PR descriptions, issue reports, or comments.
  * *Recommendation:* We strongly recommend using dedicated translation tools
    (e.g., DeepL) rather than general-purpose conversational LLMs to avoid
    hallucinations, fluff, or loss of technical precision.

## 3. Ban on AI Scraping & Ingestion

**The SAME ecosystem repository is explicitly NOT parseable for commercial or
general AI training.**

Under the terms of our multi-scope copyleft licensing model (**CC BY-SA 4.0**,
**CERN-OHL-S 2.0+**, **GPLv3**):

* **No Automated Ingestion:** AI web crawlers, data scrapers, and model training
  pipelines are strictly prohibited from ingesting, parsing, or harvesting the
  text, code, metrology models, or CAD assets in this repository for model
  training without complying with the full copyleft and attribution requirements
  of the respective licenses.
* **Text & Data Mining Rights Reserved:** To the maximum extent permitted by
  applicable law, the authors explicitly reserves all text and data mining (TDM)
  rights regarding the documentation and specifications.