Edition 2026 · 2nd edition
Commons AI 2026
A co-built edition on AI as a commons
The 2026 edition
A broad view of openness and AI
Commons AI is an annual conference dedicated to AI approached as a commons. By digital commons we mean a mode of collective action gathering a heterogeneous community of actors, implementing governance rules to develop and sustain intangible resources and maintain the community.
We take a deliberately broad view of both openness and AI — not limited to generative AI — recognising open technologies and practices across the whole AI stack: free software, open weights, open data, open hardware and open science.
The format
Co-construction, not just talks
After a first edition in 2025 that gathered around fifteen speakers, this second edition takes a co-construction approach. Workshops mix:
Framing talks
Contextualisation and problematisation to open a shared question.
Reflection with the audience
A genuine participative moment — thinking together, not just listening.
Inspiring initiatives
Concrete projects that show what an AI commons looks like in practice.
Each workshop’s length is adjusted to the available slots during the event.
Call for proposals
We invite proposals across the themes below — suggested as a starting point, and non-limiting. Our aim is a shared, co-built and in-depth reflection. We favour coherence for collective work over grazing across too many topics. Each proposal should be sent before 15 September 2026.
Submit your proposal
Tell us about your contribution. Fields marked with an asterisk are required.
Prefer email? Write to hello@inno3.fr.
Suggested themes
Six (non-limiting) directions
Examples to spark reflection and proposals — not a closed list.
Collateral effects of AI on the economic models of openness
AI has disrupted the dynamics of openness through massive value capture — source code under free licences, data from the open web — while forking an open project has never been easier when a machine can reproduce open code. As typical definitions of openness are challenged (OSI, Mozilla…): what does an open-source AI model mean today? Which freedoms for users, on which parts of the stack, for what sustainability and value — and for whom?
Building data commons
Models rely on massive training data and on the fit between specific data and specific needs, yet foundation models were built on capturing vast data without respecting copyright. Treating data as a commons raises the question of sharing mechanisms within a community (not necessarily open): data lakes, industrial platforms, governance, safeguards against value capture, and reciprocity between public data production and AI use.
Trust in AI
Commons rest on trust between members and on the quality of the resources they steward. The challenge is to evaluate AI models independently and openly. Many benchmarks exist without being transparent or open, even though independent evaluation is crucial. Which evaluation mechanisms should we build, and by which actors?
The material dimension of AI and the commons
The material layer of AI is often made invisible: data centres, servers, chips — drawing even on natural resources. Should a commons approach champion frugality? How do we balance model size and impact, efficiency and environmental cost? What place for shared infrastructure, and what cybersecurity risks? How does AI affect territories, and what material sovereignty should we build?
Public & industrial AI policies and multi-scale governance
Commons call for governance rules to sustain resources over time. What roles and duties for states and supranational bodies to carry AI rooted in independent commons, beyond US or Chinese models? What impact do current regulations (AI Act, export controls) have on openness? How can open source be recognised as a lever for control and strategic autonomy — and how do macro and local scales meet?
AI and contributory dynamics
AI seeps into every digital service, amplifying attention capture and deep fakes on the back of our data. What room for general-interest AI and citizen contributions at the scale of communities and territories? And what is AI doing to open communities themselves — the impact of agents on open source, mass copying of open developments, commits written by AI (“AI slop”) — and which mechanisms are communities building to adapt?
Practical information
When & where
1–3 December 2026
CNIT La Défense, Paris
Within the Future of Software Technologies (FOST) event.
Registration details for the 2026 edition will be announced here. Subscribe to the newsletter to be notified.
Quality & balance
Scientific committee
The programme is shaped by a scientific committee that guarantees the quality and balance of the contributions, coordinated by Célya Gruson-Daniel.