erplibre/AI_POLICY.md
Mathieu Benoit e5582e2427 [ADD] doc: adopt the OCA generative AI policy
ERPLibre had no position on AI-assisted contribution while its own history
already carried AI trailers — and carried them in the one field the OCA
policy forbids, Co-authored-by, whose meaning for a machine is legally
undefined.

The policy is adopted rather than reinvented: the file summarises it and
points at the OCA document as the reference, so it cannot drift from
upstream. What it keeps is what a contributor here needs in front of them —
the Assisted-by trailer with one line per model, the ban on AI in authorship
fields, the size and pace expectations, and the fact that responsibility does
not transfer to a tool.

Written as AI_POLICY.base.md, like README, CHANGELOG and TODO: « make
doc_markdown » globs every *.base.md, so a plain .md here would be
overwritten or, worse, silently left out of the bilingual set.

--- FR ---

ERPLibre n'avait aucune position sur la contribution assistée par IA, alors
que son propre historique portait déjà des trailers d'IA — et les portait
dans le seul champ que la politique de l'OCA interdit, Co-authored-by, dont
le sens pour une machine est juridiquement indéfini.

La politique est adoptée plutôt que réinventée : le fichier la résume et
désigne le document de l'OCA comme référence, si bien qu'il ne peut pas
diverger de l'amont. Il retient ce qu'un contributeur d'ici doit avoir sous
les yeux : le trailer Assisted-by, une ligne par modèle, l'interdiction de
l'IA dans les champs de paternité, les attentes de taille et de rythme, et le
fait que la responsabilité ne se transfère pas à un outil.

Écrit en AI_POLICY.base.md, comme README, CHANGELOG et TODO : « make
doc_markdown » balaie tous les *.base.md, si bien qu'un simple .md ici serait
écrasé ou, pire, silencieusement laissé hors de la paire bilingue.

Source: https://github.com/dixmit/oca.github/blob/ai_policy/AI_POLICY.md
Assisted-by: Claude Opus 5
2026-08-07 01:16:20 -04:00

66 lines
2.2 KiB
Markdown
Raw Blame History

This file contains ambiguous Unicode characters

This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

# Generative AI / LLM Policy
ERPLibre adopts the [OCA Generative AI / LLM
Policy](https://github.com/dixmit/oca.github/blob/ai_policy/AI_POLICY.md).
The text below is a summary; the OCA document is the reference.
## The short version
- Using AI tools to help you is fine.
- Handing over responsibility to them is not.
- Every contribution comes from a human who understands it and answers for
it, however it was produced.
- Any AI involvement means an `Assisted-by:` trailer. It is binary: either
there was AI involvement or there was not, with no threshold to judge.
- AI tools never go in `Co-authored-by:`.
- Unsupervised agentic tools are not permitted.
- If you cannot explain and defend every line, do not submit it.
- During review, engage with the feedback. Regenerating and resubmitting is
not an answer, and neither is "the AI wrote it".
- Do not post AI-generated review comments or summaries you have not
fact-checked yourself.
## Declaring AI use
Add one `Assisted-by:` line per model, in the same shape as
`Co-authored-by:`, with no blank line between them:
```text
Assisted-by: Claude Opus 4.6
Assisted-by: GitHub Copilot:gpt-5
```
The trailer says nothing about the quality of the work. It applies to every
level of use, from a piece of advice to fully autonomous coding.
`Co-authored-by:` must not name an AI tool: authorship of a work by a machine
is legally undefined. Disclosure is expected and welcome; it does not reduce
the contributor's responsibility one bit.
## Size and pace
Reviewer burden is roughly *quantity × rate*. A patch under 30 lines in a
single file is the reference point. A contribution over 500 lines needs prior
agreement with a maintainer. Contribute at a pace and size a volunteer can
actually absorb.
## Scope
This policy covers contributions to ERPLibre. Anything ERPLibre sends
upstream to the OCA is governed directly by the OCA document, including its
metrics framework and its consequences.
## Credits
Adapted from the OCA policy, itself based on the policy of the *attrs*
project. The OCA document was led by Stuart J Mackintosh, with significant
contribution from Enric Tobella Alomar, and reviewed by the OCA Governance
Working Group.