AI is tackling the great mathematical problems, writing software and acting on our behalf. It is gaining autonomy. Are we losing ours? A proof we cannot verify, a decision we cannot understand, an agent we cannot correct: what do we still control?
Choosing what we entrust to AI, understanding and challenging what it produces, deciding who it benefits: these choices must remain ours. Yet a handful of companies build the systems to which we delegate ever more. How far will we go in handing them the keys to our work, our knowledge and our decisions?
What if tomorrow’s artificial intelligence were not the preserve of a few tech giants, but the fruit of open, collective collaboration?
That is precisely Ebiose’s ambition: to let everyone take part in the AI revolution and steer its course. To choose together the challenges it takes on, the way it works and who it benefits. To stop enduring this transformation and become its actors. AI for all and by all.
AI today: a tool of control?
The loss of control begins with those who own the infrastructure. It extends to the systems we can neither examine nor modify, and on to the answers we accept without questioning them. AI already shapes how we work, learn and decide. Who gets to steer how it works: those who own it, or everyone it affects?
The AI grab: the new tech aristocracy and the computational divide
On September 29, 2026, the AI giants have lunch at the White House. The same day, an executive order imposes “super-intelligence” in place of “artificial intelligence” across the federal administration. A few days later, the president brands anyone still using the old term “the enemy”.
The episode raises a smile. Yet it says what matters. A new technological aristocracy sits at the table of power, where even the vocabulary of the debate is decided. The world’s cognitive infrastructure is concentrating in the hands of a few American players and their Chinese counterparts, who set its rules, its access and its prices.
This grip rests on a material resource: computing power. With Stargate, announced in January 2025, OpenAI and its partners aim for 500 billion dollars of investment in data centers. A computational divide is widening between those who own the machines and those who depend on them to innovate.
AI: our thoughts under influence?
The AI giants do not merely concentrate wealth. They shape what we read, the answers we receive, what strikes us as “true”, “relevant” and “desirable”. The more we delegate our judgment to them, the further their power extends.
One story accompanies this grip: AI is said to be so powerful, so dangerous, that its keys should be left with those who own it. Fear becomes an invitation to give up our power to act.
As early as GPT‑4, warnings of a possible extinction of humanity came alongside calls to regulate the most powerful models. With Astra and Fable, the implicit message remains the same: admire our power, fear its consequences, trust us. The models change. The marketing of fear starts over.
This narrative is very convenient. It sells power while legitimizing its confiscation. It passes off restricted access as prudence and sets up regulation as a barrier to competition. Those who promise to protect us consolidate their own power along the way.
In this logic, even a security flaw becomes a selling point. The Hugging Face episode offers a rather comical illustration. After the platform was hacked by its agents, one would expect OpenAI to keep a low profile. Its statement nonetheless touts the power of its models and invites professionals to try them. An isolation flaw turned into a marketing argument: that took nerve.
Being safe does not mean being protected from AI by those who own it. It means staying in control of it. We must be able to question these systems, challenge their answers and have a say in their direction. To make what AI discovers our own and build on it in turn. Progress is also measured by the power to act it gives back to us.
Tomorrow’s AI: a civic and collective alternative
Faced with this situation, it is urgent that civil society get involved in steering tomorrow’s AI and reconnect with the spirit of the Enlightenment: everyone should be able to take part in developing an AI conceived as a common good, in the service of knowledge, creativity and the great challenges of our time.
Choosing together what AI should accomplish
To democratize AI, everyone must be able to take part in its development, regardless of social background, origin, education or technical skill. The knowledge of a teacher, a nurse or a farmer must be able to steer the problems it tackles. The diversity of needs and aspirations must find its place in the choices that shape AI.
What is at stake is deciding collectively how it is used and where it is headed. Everyone must be able to propose goals, debate priorities, judge results and challenge or refuse certain uses. Choosing the problems to solve, the criteria for success and what we agree to delegate must stay in our hands. Defining what we want to improve is already a contribution to progress.
Progress everyone can build on
Progress is built by making discoveries accessible, verifiable and reusable. A mathematical proof, a method or a program take their full value when others can examine them, improve them and use them to go further. AI must extend this collective construction of knowledge.
Every advance made by AI should leave us solid foundations: knowledge and tools whose reliability we can test and whose limits we can know. By combining them, we can tackle ever more complex problems without starting from scratch.
This collective construction of knowledge must also preserve our capacity for judgment. AI’s results must be intelligible and usable by those who will rely on them. Explainability must let us understand their scope, examine their foundations, identify their limits and decide what comes next.
That is also the role of levels of abstraction: making complex work accessible to our reasoning. Bringing out the conclusions, assumptions and choices that matter, while letting us trace back to the elements that justify them. AI must let us think and act at another scale, while keeping the means to exercise our critical judgment.
This mastery is built together. Everyone must be able to contribute at the level they understand, draw on the skills of others and pass on what they have learned. Every advance in AI must increase our collective capacity to understand, to create and to act.
A collective power for the benefit of all
Designed by all, AI must also benefit all. For everyone to truly take part, the knowledge, tools and computing power required must be made accessible. Sharing the power to act also means sharing the means to act.
The rise of autonomous agents opens the way to a new digital workforce. Who will benefit from the time saved and the wealth produced? We want this power to give us more freedom to learn, create, care for ourselves and for others, and take on the challenges of our time. The gains it allows must improve our living conditions, and the value created must be shared fairly, recognizing each person’s contribution.
An AI in the service of all must also preserve everyone’s freedom and safety and take its environmental cost into account. That means weighing the expected benefits against the resources consumed, pooling means and reusing what has already been achieved. Progress must be measured by what it brings to our lives and to the world we share.
Co-building tomorrow’s AI: the Ebiose vision
With Ebiose, we want to offer the technological foundation of a democratic AI, one that everyone can help steer, build and evolve.
At the heart of this ambition is the Genomic Engine: an engine inspired by natural evolution, designed to explore, combine and test different ways of building AI. We choose the challenges; AI helps us explore the paths to meet them.
Forges and ecosystems to bring our shared ambitions to life
A forge starts from a problem we choose to solve. There, architect AIs create, test and improve agents according to the goals and criteria we define. We can then examine the results, revise their requirements and steer the rest of the exploration.
Around these forges, we want to grow collaborative ecosystems united by shared challenges: medical research, environmental protection, education. Everyone must be able to bring their knowledge, their needs or their critical eye. Field experience can open a new line of research.
A shared genome to widen what is possible
Every discovery must be able to open the way to the next ones. A verification method, a tool, a way of making agents cooperate: so many building blocks to pass on, combine and transform. Our evolution engine is designed to improve these blocks, bring out new ones and discover new ways of organizing them.
We want to gather these achievements in a shared genome: a common heritage that everyone can seize and enrich. Tested in new contexts, these blocks will be able to combine into more elaborate capabilities, which will in turn become the foundations of new discoveries. What one generation has learned to build, the next can use to go further.
This ambition extends from agents and their organization to the structure and training of the models themselves. Build ever higher, while keeping access to the foundations and the power to transform them.
An AI that evolves, control that endures
We want systems whose organization can be examined, whose transformations can be followed and whose results can be understood. Explainability must give us a grip on what AI produces: access to the assumptions, to the elements that support a conclusion and to the known limits, so we can exercise our judgment and decide what comes next.
We want to be able to correct the systems, challenge an evolution and decide what we agree to delegate. Gaining power must go hand in hand with the means to stay in control.
Making security a collective strength
Security must be built with the agents, from their design and at every evolution. That requires testing their behavior in isolated environments, limiting their access and their actions, protecting data, and being able to monitor and interrupt them. New capabilities will have to be evaluated together with the risks they introduce.
We also want to open forges dedicated to cyberdefense, to produce agents that help spot vulnerabilities, detect attacks and strengthen protections. Proven defense methods and building blocks will be able to enrich the shared genome, while preserving the confidentiality of the systems they protect. Associations, companies, local authorities and citizens must be able to benefit from these advances. The ability to defend oneself must be accessible to all.
Pooled computing power
In time, we want to rely on a peer-to-peer distributed computing infrastructure, bringing together the resources of our computers and of the other machines each of us chooses to make available. This shared power must give communities the means to pursue their own goals and widen access to both the creation and the use of AI.
We also want to make frugality a criterion of progress: reusing what has been achieved, pooling means and selecting solutions with their consumption in mind. Doing better with the resources we share.
An open economy to share the value
Compute, code, expertise, data: this project rests on many kinds of contribution. We want an economy that recognizes them, makes it possible to reward them and supports a fair sharing of the value created.
In time, an exchange marketplace, possibly backed by a currency of the ecosystem’s own, could facilitate these contributions and access to resources. Its rules should be open to discussion and steered by those who keep it alive. This economy must let everyone contribute, be recognized and benefit from shared progress.
Together, let’s build a new path for tomorrow’s AI
Ebiose is our proposal for meeting the challenges of tomorrow’s AI. To make this vision real, we need you! The democratic and responsible AI we have imagined is not ours; it belongs to all of us, and being able to build it together is a tremendous opportunity.
Test it, comment, share your ideas and remarks, let’s talk!
This is no time for resignation, but for boldness: AI can just as easily become the weapon of a surveillance-capitalism dystopia as the foundation of a democratic renaissance. Let us build now, together, the foundations of tomorrow’s AI: a democratic AI in the service of humanity.