Preamble
No one invents alone. Every scientific advance stands on the shoulders of giants. Each discovery rests on centuries of shared work: the questions others asked, the methods they refined, the knowledge they chose to publish.
The image is nearly a thousand years old. Around 1120, Bernard of Chartres said that we are like dwarfs on the shoulders of giants: we see further not because our sight is sharper, but because the giants lift us up.
Artificial intelligence is the purest expression of this truth. These systems are built from the written, scientific and artistic output of humanity itself. They exist only because generations of people thought, wrote, painted, coded and shared.
Yet today, a handful of companies hold the most powerful of these systems as private property. They decide who may use them, at what price and on what terms. They train on our collective work without asking, disclose little about how, and power their data centres with energy our planet can ill afford. The inheritance of all is becoming the estate of a few.
We are not alone in saying this. Researchers, foundations and governments already call for a public option for AI, open models and public computing infrastructure. We go one step further: the frontier models themselves must belong to everyone.
This manifesto states what follows from that, and what we must demand.
Principles
I. Progress builds on what came before
Great advances rarely come from nothing. They build on the work of those who came before. Whatever an individual or a company adds, it adds to a foundation laid by many others. Honesty about progress begins with acknowledging that debt.
II. AI is built from the commons
Today's frontier AI models, the large language models at the cutting edge, are trained on the collective knowledge, science and art of humankind. Their core methods come from open research, published and shared freely. The contributions of any single AI company, however clever, are minor compared to the foundation they build on.
This manifesto concerns these frontier models. By frontier models we mean general-purpose AI models trained on large-scale data taken from the public web and humanity's published works, at a scale of computing power only a handful of organisations can afford. Anyone may train a model solely on their own data and own it. But no one can build a frontier model that way: its capabilities exist only because it learned from all of us.
III. What is built from the commons belongs to the commons
AI models created from humanity's shared knowledge must be the shared property of humanity. No company may claim exclusive copyright or property rights over them. They must be open to study, use and improvement by everyone.
Every person, everywhere, holds four freedoms over every frontier model:
- To use it, for any lawful purpose.
- To study it: its weights, its methods and the sources it learned from.
- To improve it and adapt it to their own language, culture and needs.
- To share it and their improvements with others.
IV. AI must serve humanity, not enrich a few
AI must be used to advance science, health, education and wellbeing for all people. It must not become the private estate of a handful of billionaires. We refuse a new feudalism in which the many rent access to their own inheritance from the few.
V. AI must be ethical
Every AI system must meet three conditions:
- Ethically trained. Its training respects the people whose work it learns from, with transparency about its sources and attribution of the creators whose work shaped it.
- Lawful. It respects the laws of the country in which it is used.
- Clean energy. It is trained and run on renewable energy. A technology meant to secure humanity's future must not burn through that future to exist.
VI. A fair path from here
We do not call for seizing what exists overnight. We call for a transition: every frontier model is released openly to the public steward within a few years of its launch, and the steward may buy existing models with funds from the automation levy. Private companies remain free to build products and services on top of the common models.
Answering the objections
"Without ownership, no one will pay for AI."
Those who profit from AI should fund it. We call for an automation levy: a fee paid by companies that replace human labour with AI, or that gain economies of scale from it. This applies above all where work is done by autonomous robots powered by AI models. The gains of automation flow back into the commons that made it possible.
"Open AI is unsafe AI."
The safety measures in use today do not depend on profit. They can be applied just as well by a public steward. The most powerful models should be run by an independent, non-profit organisation under the oversight of the United Nations. This way no single company and no single nation state can turn them into a weapon.
Shared ownership means shared responsibility. The steward's governance must be independent of any one government, with its board, safety reviews and decisions open to public scrutiny.
"You cannot demand ethical training while denying ownership."
AI is already trained on humanity's knowledge without paying royalties. That is only defensible if the result returns to humanity. Taking from everyone for the profit of a few is the real injustice. Taking from everyone for the benefit of everyone, through a non-profit that belongs to all, is fair.
Call to action
None of this will be given to us. Humanity must claim it. We demand:
- Public goods. Frontier AI models are recognised in law as the common property of humanity.
- An automation levy. Those who profit from replacing human labour with AI fund the commons.
- A global steward. An independent, non-profit organisation under UN oversight runs and safeguards the most powerful models.
- Transparency. Every frontier model discloses the sources it was trained on.
- Clean energy. Every AI model is trained and run on renewable energy.
We refuse to accept that the future belongs to whoever owns the largest data centre.
We stand on the shoulders of giants. So does every machine we build. Let us make sure it carries all of us forward, not just a few.
Sign it. Shape it.
This manifesto belongs to everyone who stands behind it.
- Sign it if you share these principles. Every signature adds weight to the demands above.
- Shape it. This is a living document, like the knowledge it defends. Send us your criticism, ideas and improvements. Each new version will record what changed and why.
Version 0.1 · first draft for public feedback