Manifesto

I. Artificial intelligence belongs to everyone

AI is not a product. It is a civilizational transformation comparable in scope to the printing press or electricity. As such, it cannot be left to market forces alone. ICIA affirms that every person, every organization, every territory has the right to access the knowledge, tools and debates that shape this transformation. The UNESCO Recommendation on the Ethics of AI, adopted unanimously by 193 member states in 2021, states it unequivocally: the protection of human dignity and equity of access are the cornerstones of any serious ethical framework. Access is not a privilege. It is a condition of democracy.

II. Digital sovereignty is a condition of freedom

Delegating one’s entire cognitive infrastructure to foreign, opaque and unregulated systems is to abdicate part of one’s sovereignty. ICIA defends a European approach to AI based on transparent training data, knowledge of biases, and control over deployment conditions. The European AI Act (EU Regulation 2024/1689), the world’s first comprehensive legislative framework on AI, establishes this principle in positive law: trust is not decreed, it is organized. Not through narrow nationalism, but because a tool we do not understand is a tool we become dependent on, and unchosen dependency is a form of servitude.

III. Collective is a method, not a posture

The word « collective » in our name is not decorative. It describes an epistemology: we believe that relevant knowledge about AI is not built in the solitude of a research office, nor in the verticality of a consulting firm. It is built in the productive friction between disciplines, between professions, between generations, between uses. The OECD Principles on Artificial Intelligence, revised in 2024, confirm this: trustworthy AI requires multi-stakeholder, adaptive governance open to all actors in society. ICIA is a space of rigorous and benevolent intellectual confrontation, where the engineer, the teacher, the lawyer, the artist and the entrepreneur learn together what none of them could learn alone.

IV. Pedagogy is a political act

Training means choosing what people will be capable of doing and refusing. The European AI Act imposes, in Article 4, an AI literacy obligation for all actors who develop or deploy systems. But AI training limited to tool use without training in critical judgment, ethics and understanding of mechanisms produces docile users, not enlightened citizens. The UNESCO Recommendation states it forcefully: AI education must promote autonomy, critical thinking and fundamental rights at all stages of the system lifecycle. ICIA considers that every pedagogical action is a position on the world we want to build. We choose education that emancipates.

V. The concrete takes precedence over speculation

ICIA is not a think tank. We do not write reports that remain reports. The OECD Employment Outlook 2023 clearly establishes: 27% of jobs in OECD countries correspond to professions highly exposed to automation risk. This figure is not a theoretical abstraction. It refers to real people, in real organizations, facing real decisions. Our compass is not the theoretical state of the art, but the question of what works, why, under what conditions, and for whom. Field rigor is at least as demanding as academic rigor.

VI. Intellectual craftsmanship vs. mass production

There is an industrial way of deploying AI: replicating the same solutions in different contexts without adapting them, selling certifications without building skills, producing content without producing thought. Kate Crawford, researcher at New York University, has documented this rigorously in her book Atlas of AI: AI is not neutral, it is the product of human choices, power relations and specific contexts that determine its real effects. ICIA firmly opposes any context-indifferent approach. Every intervention, every program, every tool we design is thought for its specific terrain. Intellectual craftsmanship is not slowness: it is a requirement of relevance.

VII. Transparency is our contract with society

As a French non-profit association, ICIA has no shareholders to satisfy, no stock market valuation to defend, no interest in hiding its conclusions to protect a client. The European AI Act makes transparency a legal obligation for systems that interact with citizens. ICIA goes further: we apply this principle not only to our tools but to our organization itself. We publish what we learn. We document our mistakes as much as our successes. Our associative model is not an administrative detail. It is the structural guarantee of our independence. Transparency is not an optional virtue: it is the foundation of trust.

VIII. Human-centered is not a slogan

« Human-centered AI » has become one of the emptiest phrases in contemporary technology discourse. The Ethics Guidelines for Trustworthy AI, published by the European Commission’s High-Level Expert Group, provide a rigorous operational definition: seven concrete requirements, including human oversight, technical robustness, and non-discrimination. ICIA refuses to make it a marketing argument. Placing humans at the center concretely means: designing systems whose effects on people are anticipated, measured and assumed; refusing automations that degrade work without creating value; recognizing that AI amplifies inequalities as much as it can reduce them. This is not a humanist posture. It is a work program.

IX. Territories are not secondary markets

AI transformation cannot concentrate in a few urban hubs and leave intermediate territories without resources. The barometer published by Maire-Info in 2025 quantifies it bluntly: 59% of residents in the Paris metropolitan area use generative AI, compared to 34% in rural municipalities, creating what its authors call a « multi-speed AI France ». The Senate report on digital cohesion in territories draws the same conclusion on structural digital exclusion and calls for it to become a national priority with sustained funding. ICIA has an explicit territorial vocation. We work where resources are lacking, not only where they abound. Geographic inclusion is a full dimension of our mission.

X. Build without destroying: AI as a civilizational choice

AI can accelerate wealth concentration, weaken social bonds, automate manipulation, extend surveillance. It can also augment human capabilities, reduce inequalities in access to expertise, free up time for what has value. These two trajectories are not equiprobable. Yoshua Bengio, 2018 Turing Award laureate and one of the most rigorous scientific voices on this subject, stated it in the Journal of Democracy: the consequences of AI for democracy and humanity result from political, economic and pedagogical decisions that human actors make every day. ICIA chooses its trajectory. We build uses that strengthen individuals, collectives and institutions rather than weakening them. This is not naive optimism. It is a responsibility we assume.

Note on sources

Each link refers to a primary or institutional source: official texts of the European Commission, OECD, UNESCO, French Senate parliamentary work, published academic research. No partisan sources, no blogs, no opinion media.

Institut Collectif de l'Intelligence Artificielle, French non-profit association