Commons AI conference 10 december at La Défense

Save the date : Commons AI Conference 2025

28 octobre 2025

A community based approach for AI

The Commons AI conference – a community based approach for AI – is part of the Future of Software Technologies Conference, which will be held on the 9, 10, 11 December 2025 at CNIT FOREST La Defense.

The increased use of generative AI has brought to light areas of fragility and tension within our digital societies, such as the concentration of power, threats to personal data and the homogenisation of knowledge.

In light of the risks posed by the dominance of a few AI models and the dangers of exploiting freely accessible web resources, there is a growing consensus within open-source communities and digital commons to address these issues and develop AI that respects the resources used, the communities involved and end users.

There have been numerous initiatives at different levels, such as setting up infrastructure to host open-source AI projects, creating an LLM Open Source, using open training data, reflecting on data governance in commons approaches, and calling for regulation.

The aim of this inaugural Commons AI conference is to review initiatives that aim to develop AI using an open, commons-based approach. We will explore the findings that have already been made and consider how to consolidate and pool these initiatives.

Commons AI Program

Three sessions will be offered throughout the day, each examining one of the three main components of a common: resources, community and governance.
Bringing together different profiles (businesses, researchers, NGO and the public sector) at this meeting will facilitate an exchange of views, present the current status of ongoing projects and collectively address and identify existing and new issues.

Session 1 – Ressources to produce open, ethical, and inclusive AI

What are the resource needs of communities to produce open, ethical, and inclusive AI? How can different degrees of openness be articulated according to the components of AI (data, models and algorithms)? What open and sustainable infrastructures are available to support these projects?

11:20 AM – 11:40 AM
OpenLLM France: Building transparent and open AI with a French twist
Julie Hunter, Senior Researcher, R&D Team, Linagora

11:40 AM – 12:00 PM
Open data flows: rethinking AI infrastructure after the synthetic turn
Pierre-Carl Langlais, Co-founder, Pleias

12:00 PM – 12:20 PM
Data Spaces and Digital Commons: Building a Responsible, Transparent, and Inclusive AI Market
Bertrand Monthubert, President, Ekitia
Pauline Zordan, Lawyer, Ekitia

12:20 PM – 12:40 PM
What does it take to build effective AI systems for mapping that improve the environment?
Bertrand Pailhes, Head of Data Mapping, IGN

12:40 PM – 12:50 PM
Round Table Discussion
Facilitator: Ramya Chandrasekhar, Researcher, CNRS

Session 2 – Governance and regulatory mechanisms

What governance elements have been implemented, or need to be established, to help develop these communities and promote the implementation of these AI models using a commons approach? What regulatory mechanisms are available, whether regulatory or community-based (licences, charters, etc.)?

2:00 PM – 2:20 PM
Making data commons a prerequisite for the use of artificial intelligence
Jean Cattan, Responsable de la démarche nationale Café IA

2:25 PM – 2:45PM
Legal frictions for the re-use of the open web for AI training
Ramya Chandrasekhar, Researcher, CNRS

2:50 PM – 3:10 PM
AI Data governance: The fiduciary model as a path to collective and controlled value Creation
Vincent Bachelet, Université Paris 1 Panthéon Sorbonne, Inno3

3:15 PM – 3:25 PM
Probal.ai’s governance and revenue model: a critical open-source AI commons
Yann Lechelle, CEO, Probabl

3:40 PM – 4:00 PM
Round Table Discussion
Facilitator: Benjamin Jean, Founder and President, Inno3

Session 3 – Communities, contribution and dialogue

How are communities organised to bring together diverse actors in order to support and maintain AI as a commons? What contributions does each member make? What economic balances have been found? How can different profiles (e.g. developers, data curators and users) be brought into dialogue?

4:30 PM – 4:50 PM
When communities and industry cooperate: towards sustainable AI
Jean-Baptiste Kempf, President, VideoLAN / CEO Kyber / Tech Fellow Scaleway

4:55 PM – 5:55 PM
Une IA d’intérêt général indépendante des GAFAM : quels rôles pour les communautés ? (in French)
Round table featuring:

Jean-Marc Borredon, Director of Communication, Ville d’Annemasse
Raphaël Bournhonesque, Machine Learning Engineer, Open Food Facts
Jeanne Brétecher, Founder and Director, Social Good Accelerator
Pierre-Yves Gosset, Digital Services Coordinator, Framasoft
Jean-Philippe Clément, Deputy Director General, Ville de Paris

A summary document outlining the key points presented and some concrete avenues for collective action will be produced following this day

Coordination : Celya Gruson Daniel and Benjamin Jean (Inno3)

Speakers & Presentations

Session 1 – Ressources to produce open, ethical, and inclusive AI

Pierre-Carl Langlais – Co-founder, Pleias

Presentation: Open data flows: rethinking AI infrastructure after the synthetic turn

Abstract: Since GPT-3, language models have been trained on large corpus of unstructured web archives. Yet, this approach raises significant questions. This talk takes advantage of a recent release from Pleias, the first open and generalist synthetic environment for AI training, to explore new paradigms for AI infrastructure.

Bio: Pierre-Carl Langlais is an artificial intelligence researcher and co-founder of Pleias. A long-time advocate for open science, Pierre-Carl is an administrator on Wikipedia and co-authored an influential report for the European Commission on non-commercial open access publishing. He coordinated the publication of the Common Corpus, the largest open dataset available for training language models.

Julie Hunter – Senior Researcher, R&D Team, Linagora
Presentation: OpenLLM France: Building transparent and open AI with a French twist

Abstract: Central to our philosophy is to share all of our training data and code for data preparation and training, in addition to model weights and intermediate checkpoints—all in line with the most demanding open source requirements and ethical constraints imposed by the AI Act. In this talk, I present our philosophy of transparency, motivations and efforts to build, from scratch, large language models with a focus on French and contextualize our work in the context of other recent open initiatives.

Bio: Julie Hunter joined LINAGORA’s R&D team in 2019 to contribute linguistic expertise on corpus-centered projects. She is currently the head of research for the R&D teams projects on Natural Language Processing. In particular, she is currently the technical coordinator for the ongoing research project OpenLLM France, funded by the BPI, which seeks to develop fully transparent, open-source language models with an emphasis on the French language. Her current research questions evolve around multilinguality in language models.

Bertrand Monthubert – President, Ekitia and Pauline Zordan – Lawyer, Ekitia
Presentation: Data Spaces and Digital Commons: Building a Responsible, Transparent, and Inclusive AI Market

Abstract: Data Spaces and digital commons can provide a framework for building a responsible and inclusive AI market. By combining shared governance, interoperability, and data quality, we can develop ethical and effective AI, while integrating all stakeholders, including SMEs and citizens. However, their success depends on the ability to simplify processes, reduce costs, and build trust. To achieve this, clear rules must be co-constructed, investment made in user support, and the value of Data Spaces demonstrated through concrete projects. The goal is to create an ecosystem where technological innovation serves the public interest, based on open and collaborative infrastructures.

Bio: Professionally, Bertrand Monthubert is Professor of Mathematics at the University of Toulouse, of which he was President. He is active in the data economy and is also President of OpenIG, the regional platform for geographic information in Occitania, and President of the National Council for Geographic Information under the French government. He is an actor in the development of the use of data at territorial and national level, at the service of citizens’ uses and public policies.

Bio: Pauline Zordan is a lawyer specialising in digital law and an expert in data governance in collaborative processing between different stakeholders. For the past three years, she has been working on the concept of trusted data spaces to promote digital commons for the benefit of trusted AI.

Bertrand Pailhes – Head of Data, Mapping and Forestry operations at IGN (French national geographical institute)

Presentation: What does it take to build effective AI systems for mapping that improve the environment?

Abstract: In this talk, Bertrand Pailhès will detail how IGN frontier AI systems by building up on its skills to create unique datasets and models and share them with the broader community.

Bio: Bertrand Pailhès is specialised in French digital public policies. He led the adoption of the Digital Republic Law in 2016 that enlarged open data in France. He also coordinated the national AI strategy after the Villani report in 2018 and headed technology issues at the French national data protection agency (CNIL). Since 2024, he is head of operations at IGN and aims at introducing AI for mapping..

Ramya Chandrasekhar – Researcher, CNRS (Facilitator, Session 1)

Bio: Ramya Chandrasekhar is a lawyer and researcher, and a member of the Sustainable Data Commons (SUDACO) project. SUDACO is a collaboration between the Centre for Internet and Society of the CNRS and the Open Knowledge Foundation to create legal, technical and governance tools for sustainable and fair data sharing and reuse ecosystems.

Session 2 – Governance and regulatory mechanisms

Jean Cattan – Responsable de la démarche nationale Café IA

Presentation: Making data commons a prerequisite for the use of artificial intelligence

Abstract: The call for the creation of public data commons is becoming increasingly urgent. A distant horizon that runs counter to current laws and practices, data commons as a prerequisite for data collection remain the only option for combating the impoverishing centralization of our informational and cultural environment. The question arises in the cultural environment, but it will arise tomorrow in all forms of commerce if, as OpenAI announces, but as we anticipated two years ago, we move towards agentic commerce. The governance of the stock and its access modalities cannot be determined within the framework of bilateral relations, at the risk of creating asymmetries in favor of AI providers who are in a situation of oligopoly, if not quasi-monopoly, and are becoming the only gateways to the information sought. Requiring the establishment of governance spaces dedicated to data management and determining how it is made available is becoming a necessity in order to restore a balance of power between economic, informational, and cultural actors.

Bio: More information available at https://jeancattan.fr/Bio.html. He co-chairs the data governance working group of the Global Partnership on Artificial Intelligence (GPAI).

Ramya Chandrasekhar – Researcher, CNRS

Presentation: Legal frictions for the re-use of the open web for AI training

Abstract: Training data is key to foundation AI development – particularly Generative AI models and Large Language Models (LLMs). And a significant portion of this training data comes from the open web. But despite being lauded as a digital commons, the open web is not open for all. It is difficult to ‘see’ data flows when data and content from the open web is reused to create training datasets, and as these training datasets then move through the various stages of AI development. Legal and policy initiatives for data governance in the AI context often understand data flows as stable and traceable, when in reality, data re-use is an “inherently entangled phenomenon”.

Bio: Ramya Chandrasekhar is a lawyer and researcher, and a member of the Sustainable Data Commons (SUDACO) project. SUDACO is a collaboration between the Centre for Internet and Society of the CNRS and the Open Knowledge Foundation to create legal, technical and governance tools for sustainable and fair data sharing and reuse ecosystems.

Vincent Bachelet – Université Paris 1 Panthéon Sorbonne, Inno3

Presentation: AI Data Governance: The Fiduciary Model as a Path to Collective and Controlled Value Creation

Abstract: The rapid development and democratization of AI has raised questions about the fundamental principles of free software and open source. Free licenses have long formed the basis of strategies for sharing and protecting digital resources. However, they are now insufficient for ensuring respect for the values of open source and open data. In a context where value lies more in leveraging data than in owning intellectual property rights, it is necessary to rethink data governance mechanisms to ensure data is leveraged collectively and under control. The Data Governance Act charts a promising course by encouraging collaborative models, such as data spaces and data cooperatives. These models can be compared to data trusts, which are developed in common law jurisdictions, and “fiducie de données”, which have been trialed in Quebec. In France, where there is no direct equivalent to these instruments, an innovative transposition could be achieved by creatively repurposing “fonds de dotation,” associations with legal personality and a mission of general interest. Through examples, we will see how data governance can break free from the traditional proprietary model and adopt more collective and reflexive forms adapted to the challenges posed by AI.

Bio: Vincent Bachelet, who holds a doctorate in law, is a specialist in intellectual property and digital law. He is an expert in open source law, open models, and digital commons. As a postdoctoral researcher at the University of Paris I Panthéon-Sorbonne’s IRJS, he focuses his research on open source artificial intelligence and intellectual property rights compliance. He is also a consultant at inno³, where he helps projects structure and implement their open-source strategies.

Yann Lechelle – CEO, Probabl

Presentation: Probal.ai’s governance and revenue model: a critical open-source AI commons

Abstract: In an era when AI is reshaping value chains, the issue has become deeply political and economic, not just technological. This discussion will address the governance and revenue model of Probabl.ai, a mission-driven company that operates a critical open-source AI commons. More broadly, it will address the conditions for the sustainability, impact, and scaling of digital commons. These reflections expand on the ideas presented in Ouvertarisme – Le Numérique des Lumières, which was recently published and is available on bod.fr, fnac.com, and amazon.com. There will be two book signings about the book during the Salon. Join us for an open discussion on how to reconcile public interest, value creation, and resilience in the AI economy.

Bio: Yann is co-founding CEO at Probabl on a mission to distribute open source data science and machine learning globally. Previously, he was CEO at Scaleway, a leading regional cloud infrastructure provider with unique European attributes and values. Over the past two decades, he has founded, grown and exited a total of 5 tech startups. One of these startups was Snips.ai, a leading voice processing and AI deeptech that he brought from seed to exit in a trade sale to Nasdaq-listed Sonos. Yann is also founding member of France Digitale, board member at HUB France AI and One-o-One. He gives back to the community as an angel investor, founder mentor and Entrepreneur-in-Residence at INSEAD. Yann holds a Bachelor of Computer Science summa cum laude from the American University of Paris, as well as an MBA from INSEAD.

Benjamin Jean – Founder and President, Inno3 (Facilitator, Session 2)

Bio: Founder and president at Inno3, Benjamin Jean specializes in open source law, open models, and digital commons, helping projects structure and implement their open-source strategies.

Session 3 – Communities

Jean-Baptiste Kempf – President, VideoLAN / CEO Kyber / Tech Fellow Scaleway

Presentation: When communities and industry cooperate: towards sustainable AI

Abstract: Jean-Baptiste Kempf, working at the intersection of open source communities, digital companies, and end users, will share his vision of how diverse communities—including developers, data and model experts, users, and industry professionals—can collaborate to create AI designed as a « common good. » Drawing on his experience with iconic projects such as VideoLAN and his work with digital companies, he will demonstrate how open-source models of contribution, governance, and coordination can inspire the AI ecosystem, promoting collective organization, economic balance, and the articulation of roles. Finally, he will invite us to rethink the boundaries between associations, companies, and users to create open, interoperable, and sustainably governed AI infrastructures.

Bio: Technology and Free Software enthusiast with a wide scientific background. I am an entrepreneur, a CTO, Tech advisor, Startups board member and President of an Open Source NGO, working on VLC & FFmpeg. I am one of the major developer and leader of VLC Media Player, an iconic open-source project with over 5 billion downloads. From VLC to VideoLabs, Kyber, and Playuro, my work has been focused on creating tech solutions that are always state-of-the-art but ethical, serving industries from media, gaming to cloud computing. I created several bootstrapped companies around the open source multimedia ecosystems, where the clients were the global tech companies like Meta, Google, Electronic Arts, Netflix, Thales, and TikTok, who rely on our technology to stream video on the internet. With a passion for open-source development, high-performance computing, and hardware design, I’ve had the privilege to lead and collaborate with exceptional engineering teams. I’ve been auditing, advising and leading small and large engineering organizations, for VCs, for small startups or large entreprise. Today, my focus is on new deep-tech ventures and working on new projects that push the limits of what’s possible while prioritizing sustainability and team well-being. Whether it’s creating seamless cloud gaming experiences or developing cutting-edge video solutions, I believe that technology should not only solve complex problems but also create positive change for the world.

Table ronde en français : Une IA d’intérêt général indépendante des GAFAM : quels rôles pour les communautés ?

Abstract: Une IA d’intérêt général, s’appuyant sur des communautés et des projets Open Source, souverains, frugaux, capacitants et indépendants des GAFAM, est-elle possible? Nos intervenants, acteurs majeurs et innovants de l’Open Source, de l’Open Data, de l’ESS et du secteur public, partageront leurs réalisations prometteuses, leurs espoirs et leurs doutes. Nous essaierons collectivement d’esquisser un chemin d’usages vers cette Commons AI que nous souhaitons tous.

Jean-Marc Borredon – Director of Communication, Ville d’Annemasse

Bio: Directeur de la communication à Annemasse Agglo, communauté d’agglomération, manager d’une équipe de 8 personnes et enseignant/formateur en communication publique, médias sociaux, gestion de projet. Formé en lettres modernes et à l’IA avec une double spécialité en informatique et sciences sociales. Intimement investi par les valeurs du service public et passionné par le numérique, les réseaux sociaux et la communication publique.

Raphaël Bournhonesque – Machine Learning Engineer, Open Food Facts

Bio: Diplômé de l’ENS Lyon, Raphaël Bournhonesque est actuellement ingénieur Machine Learning chez Open Food Facts. Il est spécialisé en Machine Learning, Deep Learning et traitement automatique du langage naturel (NLP).

Jeanne Brétecher – Founder and Director, Social Good Accelerator

Bio: Fondatrice et directrice de Social Good Accelerator pour le développement de l’ESS numérique #Europe #SocialTech. Fondatrice et directrice conseil de Jungle Coop: stratégies partenariales pour l’intérêt général #ESS #RSE.

Pierre-Yves Gosset – Digital Services Coordinator, Framasoft

Bio: Coordinateur des services numériques chez Framasoft.

Jean-Philippe Clément – Deputy Director General, Ville de Paris

Bio: Directeur Général adjoint Services en charge de l’Espace Public et des Equipements de proximité de la Mairie du 17e arrondissement de Paris. Jean-Philippe Clément est également animateur de l’émission « Parlez-moi d’IA » sur la radio FM-DAB+ Cause Commune (93.1FM à Paris). Ancien responsable de la mission Ville Intelligente et Durable et de la démarche data de la Ville de Paris, les sujets de la transformation et de la transition numérique et environnementale et de l’accompagnement des métiers font également partie de sa veille et sujets d’intérêt.