Skip to Content

Charter for the use of artificial intelligence by Pratiques en Santé

Daniel Oberlé - My commitments to responsible, transparent use in the service of public health
5 August 2026 by
Charter for the use of artificial intelligence by Pratiques en Santé
Daniel Oberlé - Pratiques en santé Oberlé
| No comments for now


1. Preamble

Pratiques en Santé is a public health resource platform, active since 2005. It provides prevention actors, health promotion, the social and medico-social sector with monitoring, documentary summaries, thematic files and tools to support their practices.

Artificial intelligence (AI) is now part of the means mobilised to produce and disseminate these contents: document summaries, writing aid, accessibility, illustration, audio or video summaries. This evolution calls for a clear framework. This charter explains why and how AI is used for the operation of the platform, according to which principles and with what guarantees.

This use of AI is inseparable from the values that underpin the platform: health literacy, empowerment, equity, intersectorality and the fight against discrimination. AI only makes sense if it serves these purposes — never if it weakens them.

A principle distinguishes my approach: AI never produces from nothing. The summaries and tools that it helps to create are developed solely from sources and documents selected by Daniel Oberlé, in light of over 45 years of experience in the field of public health practices and policies, prevention, the social and medico-social sector, and the daily monitoring work carried out. The AI formats a chosen corpus; it does not replace this work of selection and expertise.

In one sentence

At Pratiques en Santé, artificial intelligence is a tool of assistance. The editorial decision, verification and the responsibility for the content remain, in all circumstances, human.

2. Scope of application

This charter applies to all content designed and published by Pratiques en Santé, regardless of the medium:

  • the blog articles and the editorials of the weekly newsletter (L'Hebdo) ;

  • the Notebooks / Open Notebooks (thematic documentary files) ;

  • the Field cases (case studies, the action sheets) ;

  • the sheets in Easy to read and understand (FALC) ;

  • the videos and decoding podcasts published on the YouTube channel ;

  • the visuals, banners and infographics accompanying the publications ;

  • the English version of the platform, resulting from the automated translation of content.

She engages me as an 'editor-facilitator of the platform.

3. My five commitments

My uses of AI are based on five commitments, which guide all the practical rules detailed further on.

1. Humans decide, AI assists

No content is published without proofreading, validation and appropriation by a human. AI proposes; we dispose. The editorial responsibility for each publication is fully assumed by Pratiques en Santé.

2. Transparency

We indicate fairly the use of AI when it is substantial, so that our readers and partners know which content they are interacting with.

3. Reliability and rigor of sources

AI is not authoritative. The information, figures and references produced or reformulated with its help are systematically verified and linked to identified and reliable sources.

4. Respect for people and data

We protect personal and confidential data: no sensitive data is entrusted to an AI tool, in compliance with the GDPR.

5. Equity, accessibility and non-discrimination

Constant vigilance regarding model biases. AI is used to make information more accessible (including easy-to-read formats and various tools), never to produce or relay stigmatizing content.

4. Our practical rules

4.1 Transparency towards readers

  • Article written and structured by Daniel Oberlé with the assistance of Claude (Anthropic) for the documentary synthesis from a selected corpus, and illustrated via ChatGPT. The verification, proofreading and editorial responsibility are entirely human.

  • When a content has been produced or significantly modified with the help of AI, we clearly and accessibly indicate it (mention at the foot of the article, footnote or description associated with the video).

  • The audio or video summaries generated automatically mention the original sources, so that the reader can trace back to the document analyzed.

  • The English version of the site is produced by automated translation: it is indicated as such, and the French version is referenced in case of doubt or inaccuracy.

  • This transparency does not dilute our responsibility: it accompanies it.

4.2 Reliability and editorial responsibility

  • All content goes through a human critical proofreading before publication. AI models can produce errors or fabricated statements (“hallucinations”): nothing they generate is taken for granted.

  • The data, quotes and references are verified at the source. The sources remain distinct from the synthesis text and are explicitly cited.

  • AI is not used to provide individual medical advice or clinical recommendations : our content is aimed at professionals and the general public for information and skill enhancement purposes, and does not substitute for medical advice.

  • In the event of doubt about information produced with AI, we refrain from publishing it until it has been confirmed.

4.3 Data protection and privacy (GDPR)

  • No personal or sensitive data (health data, data identifying a person, confidential information from partners) is entered into an AI tool, regardless of the guarantees displayed by the tool's publisher.

  • All data or case studies submitted to AI tools are fictitious. No personal or sensitive data is transmitted to the models.

  • We work from public documents, anonymised or whose dissemination is authorised.

  • The privacy settings of the tools used are configured, where possible, to limit the reuse of our content for training purposes.

  • In the event of the inadvertent sharing of confidential information, the situation is corrected without delay.

4.4 Equity, accessibility and combating discrimination

  • We remain vigilant to the biases that models may reproduce (gender, origin, age, social situation, disability). The content is reviewed in light of our values of inclusion and anti-discrimination.

  • AI is put to the service of the accessibility of information, notably through the production of Easy to read and understand (FALC) sheets, to reduce inequalities in access to knowledge in health.

  • The formulations produced by AI are reviewed to avoid any stigmatising or blaming language towards individuals and vulnerable groups.

4.5 Tools and governance of uses

We use generalist AI tools, publicly available, as a user (and not a designer) of these systems. Each tool is mobilised for a specific use:

  • The tools only work from the sources that we provide to them — documents selected according to the field experience of Pratiques en Santé and its daily monitoring. The AI synthesises and formats a chosen corpus; it does not replace this selection work.


AI ToolRole in Pratiques en SantéHuman guarantee & Scope
Claude (Opus)Assistance with writing, documentary summaries, structuring of FALC sheets and case studiesExclusive processing of verified and pre-selected documentary corpus.
NotebookLMCreation of video/audio summaries for the YouTube channelVisual and audio mediation of pre-existing articles on the site.
PerplexityInformation search and cross-referencing of sourcesSourcing and systematic verification of official publications.
ChatGPT Generation of banners and visual illustrationsGraphic creation only (does not produce editorial content).

This table is evolving: the addition or replacement of a tool is decided based on the same criteria — reliability, data protection, environmental cost and alignment with the values of the platform.

5. Legal and reference framework

Our approach is framed within the European and French legal framework applicable to AI and data.

  • Regulation european on AI (AI Act, EU regulation 2024/1689). Entered into force on 1 August 2024, it applies progressively. Since 2 August 2026, the transparency obligations (Article 50) require clearly indicating content generated or substantially modified by AI, as well as interactions with a dialogue system. Our uses fall under a low to minimal risk, but we apply these transparency obligations.

  • General Data Protection Regulation (GDPR). It regulates all processing of personal data, including when it is done through an AI tool. The AI Act and GDPR are complementary and can apply together.

  • Recommendations from the CNIL. We follow the guidelines from the CNIL regarding the responsible use of AI and data protection.

This charter has no binding legal value towards third parties; it formalises our commitments and good practices.

6. Monitoring, updating and contact

  • Referent. Daniel Oberlé, facilitator-editor of Health Practices (profile: linkedin.com/in/daniel-oberle), is responsible for the application and monitoring of this charter.

  • Revision. The charter is reviewed at least once a year, and with each significant change in tools, uses or the regulatory framework. As AI technologies evolve rapidly, this document is meant to be dynamic.

  • Dialogue. This charter is open to feedback from our readers and partners: your comments help us improve it.

  • Contact: pratiquesensante@pm.me

Appendix — Reference resources

This charter is based on the following public resources, useful for exploring the topic further or building your own approach.

Regulatory framework


Practices in Health — Charter for the use of artificial intelligence — Version 1.0, 05 August 2026


in NEWS
Sign in to leave a comment