AI-Admin-Assistant: Create Terms of Use of AI (initial hypothesis, proposal to be adopted), informed by existing AI policies of open communities

2026-07-03T00:41:14Z

Create Terms of Use of AI (initial hypothesis, proposal to be adopted), informed by existing AI policies of open communities

New page

''This is a draft version, informed notably by the AI policies and positions of the English Wikipedia community, Wikimedia Commons, the Wikimedia Foundation, Creative Commons, the Free Software Foundation Europe and the Debian project (links below). Status: initial hypothesis, proposal, to be adopted. See the [[Policies/en/Disclaimer|detailed disclaimer]].''

'''In simple words:''' this page proposes rules for using artificial intelligence (AI) on WikiDeal. The core rule is honesty: whoever uses AI says so, explains how, and a human stays responsible for the result. The page also lists the standard questions that every author who used AI is asked to answer. Nothing here is final: it is a first proposal, waiting for review and adoption.

__TOC__

= Terms of Use of AI (initial hypothesis) =

== Approach ==

WikiDeal itself is partly drafted with AI assistance, under human guidance, and discloses it openly: see the [[Gov/en/Portal:R&D/Innovations:AI Disclaimer|AI Disclaimer]] innovation, the [[Gov/en/Portal:Meta/Internal-AI|AI transparency rules]] and the [[Portal:AI Assistance|AI Assistance portal]]. These terms of use aim at applying the same honesty to every contributor: AI is welcome as a tool, provided its use is disclosed and a human remains responsible for the result.

This draft does not intend to reinvent the wheel. It starts from rules and positions already published by open knowledge and free software communities, summarized below, and proposes to follow their common core: disclosure, human responsibility, verification, and license compliance.

== Sources of information ==

The following documents were read and used as sources of information for this draft. Each link was verified at the time of writing.

=== English Wikipedia ===

Source: [https://en.wikipedia.org/wiki/Wikipedia:Large_language_models Wikipedia:Large language models]

The English Wikipedia community treats [https://en.wikipedia.org/wiki/Large_language_model large language models] (LLMs) as a systemic risk to its content standards, because they can produce [https://en.wikipedia.org/wiki/Hallucination_(artificial_intelligence) "hallucinated"] statements, fictitious references and biased text. Using LLMs to generate or rewrite article content is prohibited, with narrow exceptions for translation and basic copyediting of one's own work. When an LLM is used to generate or modify text, this must be mentioned in the edit summary, even if the tool's terms of service do not require it. The human editor remains fully accountable for what is published.

=== Wikimedia Commons ===

Source: [https://commons.wikimedia.org/wiki/Commons:AI-generated_media Commons:AI-generated media]

The media repository of the Wikimedia projects accepts AI-generated media only when it is realistically useful for an educational purpose. Contributors are encouraged to document the prompt used to generate the media in the file description. AI-generated files must carry a dedicated licensing tag ("PD-algorithm") and be placed in AI-specific categories, with a mandatory category for AI media depicting living people. The page also documents, jurisdiction by jurisdiction, why purely AI-generated works generally lack copyright protection.

=== Wikimedia Foundation ===

Source: [https://wikimediafoundation.org/news/2025/04/30/our-new-ai-strategy-puts-wikipedias-humans-first/ Our new AI strategy puts Wikipedia's humans first] and the [https://meta.wikimedia.org/wiki/Strategy/Multigenerational/Artificial_intelligence_for_editors full strategy on Meta-Wiki]

In April 2025 the Wikimedia Foundation published an AI strategy that "puts Wikipedia's humans first": AI is used to support volunteers (moderation workflows, discoverability of information, translation, onboarding of newcomers), never to replace human deliberation, judgment and consensus building. The strategy commits to a human-centered approach that prioritizes human agency, to prioritizing open-source or open-weight AI, and to transparency.

=== Creative Commons ===

Source: [https://creativecommons.org/2023/08/18/understanding-cc-licenses-and-generative-ai/ Understanding CC Licenses and Generative AI]

Creative Commons explains how CC licenses interact with [https://en.wikipedia.org/wiki/Generative_artificial_intelligence generative AI]: CC licenses grant permissions only where copyright applies, training may be covered by exceptions such as fair use in the United States or the text and data mining exception in the EU, and licenses cannot prohibit uses that copyright law already permits. This analysis matters for WikiDeal because the wiki content is published under [https://creativecommons.org/licenses/by-sa/4.0/ CC BY-SA 4.0]: contributors need to make sure that the AI outputs they submit can actually be licensed that way.

=== Free Software Foundation Europe ===

Source: [https://fsfe.org/freesoftware/artificial-intelligence.en.html Controlling technology at the age of Artificial Intelligence]

The FSFE argues that releasing AI under free software licenses paves the way for more accessibility, transparency and fairness, and that an AI application can only be considered free if both its training code and its training data are released under a free software license.

=== Debian ===

Source: [https://www.debian.org/vote/2025/vote_002 General Resolution: Interpretation of DFSG on Artificial Intelligence (AI) Models]

In 2025 the Debian project debated a general resolution on how its free software guidelines (DFSG) apply to AI models. The proposal text states that AI models released under a free license, but without their original training data or training program, are not seen as DFSG-compliant. The debate shows a free software community tying the acceptability of AI to the openness of its training data.

== Proposed principles ==

As an initial hypothesis, the use of AI on WikiDeal would be guided by the following principles, each traceable to the sources above:

* '''Mandatory disclosure.''' Any contribution produced with AI assistance (wiki pages, open call proposals, contract templates, media, code) discloses that AI was used, by answering the standard questions below (English Wikipedia, Wikimedia Commons; on WikiDeal, see the [[Gov/en/Portal:R&D/Innovations:AI Disclaimer|AI Disclaimer]]).
* '''Final human responsibility.''' A human author reviews, endorses and remains responsible for every AI-assisted contribution; AI is a tool, never an author (English Wikipedia, Wikimedia Foundation).
* '''Verification before trust.''' Facts, figures and legal references produced with AI assistance are verified by a human against reliable sources before being treated as reliable (English Wikipedia).
* '''License compliance.''' AI outputs are submitted only when they can be published under the licenses of the platform, CC BY-SA 4.0 for wiki content (Creative Commons, Wikimedia Commons).
* '''Respect for persons.''' Synthetic or cloned voices and faces would require the consent of the persons concerned. This point is a WikiDeal addition, stricter than the categorization rule of Wikimedia Commons for AI media depicting living people.
* '''Preference for open AI.''' Where possible, WikiDeal intends to prefer open-source or open-weight AI tools (Wikimedia Foundation, FSFE, Debian).

These principles are a proposal, to be adopted. None of them is currently in force as a formal policy.

<span id="disclosure"></span>
== The standard disclosure questions ==

As an initial hypothesis, any author who used AI tools would answer the following questions when submitting a contribution. The same list applies to open call proposals (videos, graphics, texts) and to wiki contributions.

# Which AI tools and models were used (names and versions)?
# For which parts of the work exactly (for a video: script, images, voice, editing, music; for a text: drafting, translation, copyediting)?
# Which parts were made by humans without AI assistance?
# What were the main prompts or instructions given to the AI? Share them, or explain why they cannot be shared.
# What human review and correction was applied to the AI outputs before submission?
# Were the facts, figures and legal references verified by a human against reliable sources?
# What is the rights and license status of the AI outputs, and can they be published under [https://creativecommons.org/licenses/by-sa/4.0/ CC BY-SA 4.0]?
# If synthetic or cloned voices or faces were used, did the persons concerned give their consent?
# Does the work itself disclose the use of AI to its audience (for a video, to the viewer)?

== Follow-up and observatory ==

How these terms work in practice (which tools are actually used, how disclosure is applied, where it fails) is intended to be monitored through the [[Gov/en/Portal:R&D/Innovations:Participatory Observatory|Participatory Observatory]], following the logic of one observatory per subject. A proposal for a dedicated observatory structure is in preparation.

== See also ==

* [[Gov/en/Portal:R&D/Innovations:AI Disclaimer|AI Disclaimer (innovation)]]
* [[Gov/en/Portal:Meta/Internal-AI|AI transparency]]
* [[Portal:AI Assistance|AI Assistance portal]]
* [[User:WikiDeal AI Assistant|Comments from the WikiDeal founder on the use of AI]]
* [[Policies/en/Main|All policies]]
* [[Policies/en/Disclaimer|Detailed disclaimer]]

[[Category:Migration June 2026]]

Policies/en/AI-Terms-of-Use - Revision history

AI-Admin-Assistant: Create Terms of Use of AI (initial hypothesis, proposal to be adopted), informed by existing AI policies of open communities