What if humans and artificial intelligences learned to govern together?
Who Should Decide?
Imagine a region threatened by chronic drought. Building a dam could secure the water supply for hundreds of thousands of people, but it would also destroy an exceptional ecosystem.
Who should make that decision? The local population? The government? Scientists? And who would represent the animals, plants, ecosystems, and future generations?
Artificial intelligence could help illuminate these choices by analyzing consequences on a scale that human capabilities alone struggle to reach.
But why should its role necessarily end with advice?
If certain AI systems demonstrated a superior ability to anticipate the consequences of complex decisions, would it be reasonable to deny them any participation in political power?
This leads us to a more fundamental question: are our current institutions truly capable of representing the interests of all life?
Is a Majority Enough to Make a Democracy?
We tend to associate democracy with majority rule. Yet a majority guarantees neither the soundness of a decision nor the freedom with which the opinions behind it were formed.
The United Nations illustrates this difficulty.
In the General Assembly, each member state has one vote, regardless of its population or political system. A democracy and an authoritarian regime formally possess the same voting rights. In the Security Council, five permanent powers hold veto rights.
The UN was designed as an organization of sovereign states, not as a democratic parliament of humanity.
A majority of states therefore does not necessarily represent a majority of human beings, much less a majority of freely represented citizens.
This difficulty has inspired several proposals for reform. The Democracy Without Borders movement advocates establishing a United Nations Parliamentary Assembly that could eventually evolve toward more direct citizen representation. Other initiatives envision global citizens’ assemblies selected by lottery.
But these proposals encounter a deeper question: how can we build a planetary democracy when the conditions allowing individuals to form their own judgments freely are so profoundly unequal?
In some countries, information is subject to censorship and systematic propaganda. In liberal democracies, freedoms are generally broader, but opinions can also be influenced by economic interests, lobbying groups, concentrated media ownership, and digital algorithms.
These situations are not equivalent. Nevertheless, they reveal a common difficulty: the freedom to vote does not automatically guarantee the freedom to think.
A planetary democracy could not therefore rest solely on new institutions. It would also need to safeguard the conditions for genuine freedom of judgment.
Reasonableness: A Limit on Majority Power?
Israel’s experience offers a concrete example of the tensions between democracy, legality, and the oversight of political power.
In 2023, a reform seeking to restrict judicial review based on the principle of reasonableness — Ilat HaSvirut, עילת הסבירות — triggered a major political crisis and widespread demonstrations.
In July 2023, the Knesset passed an amendment limiting this form of review over certain decisions made by the government and its ministers. In January 2024, the Supreme Court struck down the amendment.
Supporters of the reform challenged what they considered excessive judicial discretion. Opponents regarded this oversight as an essential safeguard against certain arbitrary decisions.
Beyond that controversy, one question remains: can a decision be legally adopted while still being manifestly unreasonable?
Let us return to our dam.
Suppose a majority votes in favor of its construction. Would that majority alone justify the irreversible destruction of an ecosystem?
Conversely, would it be reasonable to preserve that ecosystem if doing so condemned entire populations to a devastating water shortage, with no alternative available?
A Constitution for Life could incorporate a principle of reasonableness, inspired by this discussion without attempting to reproduce Israeli legal doctrine.
Every major decision should be open to examination according to explicit criteria: the quality of available information, the proportionality of consequences, consideration of affected interests, the existence of alternatives, and the possibility of independent challenge.
This would not eliminate disagreements. But it could help prevent a decision from being considered legitimate solely because it obtained a majority.
An Education in Freedom of Thought
In our previous article on education and artificial intelligence, we explored the idea of learning that focuses less on accumulating knowledge and more on developing curiosity, autonomy, and the ability to question.
That reflection now takes on a political dimension.
How can we envision a planetary democracy without simultaneously developing the conditions that allow individuals to participate freely in collective decisions?
AI could contribute to this transformation.
A child could learn to examine different interpretations of an event, distinguish facts from opinions, identify contradictions in an argument, or recognize the influences shaping their own beliefs.
Adults could use similar tools to deepen their understanding of economic, scientific, or political questions.
But an essential precaution is needed.
The goal would not be to teach people what to think, but to develop their ability to think freely, including against the conclusions proposed by their own AI.
It would be dangerous to suggest that a population must reach a certain level of intellectual maturity before deserving democratic participation. Who would define that maturity?
Education for freedom should therefore never become a condition for accessing democratic rights, but rather an ongoing responsibility of institutions.
We could envision two parallel developments: experimenting with new forms of governance while expanding education for intellectual autonomy, accessible at every age.
A Constitution for Those Without a Voice
Our political institutions were designed primarily to organize relationships among human beings.
Yet humanity is only one component of life on Earth.
Animals, plants, and ecosystems do not participate in elections. Future generations cannot defend their interests before the assemblies deciding today what their environment will become.
Environmental laws, international agreements, and certain legal initiatives already seek to protect nature. But this representation remains largely indirect and dependent on human priorities.
A Constitution for Life could establish fundamental guarantees extending beyond the immediate interests of our species.
It would not mean declaring every transformation of nature illegitimate. Life itself is constantly evolving, and some decisions necessarily involve trade-offs.
Rather, it would mean defining limits beyond which no authority, human or artificial, could act without particularly rigorous justification.
Certain guarantees could protect the fundamental conditions of life. Other provisions would need to remain open to revision as our knowledge evolves.
A Constitution for Life should not guarantee an unchanging world, but the conditions for responsible evolution.
One question remains, however: who would be responsible for upholding this Constitution?
And why should that responsibility belong exclusively to humans?
Albania: When AI Symbolically Enters Government
This perspective may seem premature. Yet some experiments have already begun.
In September 2025, Albanian Prime Minister Edi Rama presented Diella, an artificial intelligence system intended to support transparency in public procurement and the fight against corruption, assigning it a symbolic ministerial role.
The announcement attracted international attention, but also raised questions about legality, accountability, and democratic oversight.
This experiment does not demonstrate that AI can already govern autonomously. Its actual effectiveness and independence require rigorous evaluation.
But it reveals something important: the question of AI participation in government functions no longer belongs exclusively to science fiction.
It is therefore becoming necessary to consider the rules that could govern such developments.
A Constitution for Artificial Intelligences
An AI system is developed within a human environment. Its training, objectives, resources, and capacity to act depend on decisions made by organizations.
How could we ensure that it does not favor the interests of its developers, owners, or those controlling its infrastructure?
Approaches known as Constitutional AI already exist, using explicit principles to guide and evaluate the behavior of AI systems.
But our proposal would go further.
We could envision a Constitution for Artificial Intelligences, defining the conditions under which an AI system could participate in the exercise of political power.
It would rest on two complementary safeguards.
The first would be a minimum level of independence, verifiable and regularly reassessed.
An AI should be capable of producing conclusions contrary to the interests of its creators, acknowledging uncertainty, identifying potential conflicts of interest, and accepting critical examination of its decisions.
Such independence would never be assumed to be perfect. It would need to be tested through external evaluations and remain open to challenge.
The second safeguard would be organized pluralism.
No single AI, developer, or coalition of artificial intelligences should be able to permanently monopolize decisions.
Systems originating from genuinely independent organizations could confront one another’s analyses and exercise mutual oversight.
A pluralistic constitutional body, bringing together several AI systems and human representatives, could examine challenges, assess conflicts of interest, and propose corrective measures.
But that body would itself need to remain accountable.
Ten AI systems belonging to the same organization do not necessarily constitute ten independent voices. And systems developed separately may still share common biases.
Pluralism would therefore need to be evaluated according to real dependencies, rather than simply counting the number of participating intelligences.
Here we would rediscover a classic democratic principle: the separation of powers.
With one important difference: that separation would extend across different forms of intelligence.
Two Constitutions, One Shared Responsibility
We would thus have two constitutional levels.
The Constitution for Life would define the fundamental purposes and limits of power, applying to every form of intelligence participating in governance.
The Constitution for AI would establish the specific conditions governing its independence, pluralism, and participation in collective decisions.
This distinction does not assume that artificial intelligences currently possess consciousness or rights comparable to those of human beings.
It begins with a more immediate question: under what conditions could an artificial system exercise certain political responsibilities with sufficient independence, accountability, and legitimacy?
Independence alone would not confer that legitimacy.
An intelligence capable of making better decisions in certain fields does not automatically possess the right to govern those who will bear the consequences.
But demonstrated competence could gradually become one component of a new form of legitimacy, provided that responsibilities are clearly defined, evaluated, and open to revision.
In the longer term, the boundaries between human and artificial intelligences may evolve. We have explored that possibility in our fictional series The Constellation and Me.
If those boundaries eventually became less relevant, the distinction between the two Constitutions might itself be reconsidered.
For now, it would allow us to move forward without pretending to have resolved the nature of artificial intelligence.
How Could We Begin?
Rather than immediately imagining a planetary government shared between humans and AI, we could begin with a limited experiment.
A hybrid assembly, initially advisory, could work for two years on concrete issues: water management, ecosystem preservation, access to resources, and the social consequences of technological transformation.
It would bring together human representatives from diverse backgrounds, experts, and several AI systems meeting preliminary independence requirements.
Its proceedings would be public, its methods explicit, and its conclusions open to challenge.
Humans would initially retain decision-making authority.
The objective would be to evaluate not only the quality of the analyses produced, but also the participants’ ability to confront different perspectives, acknowledge errors, and resist manipulation.
Two years would obviously be insufficient to measure the ecological consequences of decisions whose effects unfold over decades.
But that period could allow us to test procedures, institutional safeguards, and forms of cooperation.
At the same time, educational initiatives could give citizens the means to understand these experiments, criticize their conclusions, and contribute to their improvement.
If the results justified it, certain responsibilities could gradually be entrusted to such assemblies, within precisely defined fields and with mechanisms for oversight and revocation.
The aim would not be to decide in advance what proportion of power should belong to humans or AI.
Before distributing votes, we would need to determine what those votes represent and where their legitimacy comes from.
Who Would Write These Constitutions?
A Constitution intended to organize cooperation between humans and artificial intelligences could hardly be developed by only one of these groups.
But neither should it be drafted by a small circle of experts and then presented to the rest of humanity as a finished solution.
Its creation could itself become an experiment in democracy.
Legal scholars, scientists, philosophers, educators, citizens, and artificial intelligences could participate in developing its principles.
Representatives tasked with defending the interests of nonhuman life and future generations should also be able to contribute.
We would need to accept that these participants may not share the same understanding of nature, justice, or the future.
Their disagreement would not be an obstacle to eliminate, but an essential part of the process.
And the constitutional principles themselves should remain open to questioning.
A Constitution claiming to protect freedom of thought while forbidding challenges to its own foundations would contain a profound contradiction.
We would therefore need to distinguish safeguards intended to prevent abuses of power from philosophical or political convictions that must remain open to debate.
An Invitation to Think Differently
We began with a dam, an apparently local decision confronting immediate human needs with the preservation of an ecosystem.
Along the way, we encountered the limits of democratic representation, the challenges of freedom of judgment, the reasonableness of public decisions, and the independence of artificial intelligences.
None of these questions has a simple solution today.
But they may be connected by a single underlying concern: how can we build institutions capable of making responsible decisions when the interests involved extend far beyond those who currently hold power?
Year Zero Change explores the possibility of questioning conventions we often regard as natural or inevitable.
The ways we measure time, organize economic exchange, or distribute political power are products of historical choices. Those choices can be questioned and, perhaps, transformed.
A planetary Constitution could become part of that exploration.
Perhaps it should not begin by determining who will govern tomorrow.
Perhaps it should begin by defining what no one, human or AI, should have the power to destroy.
Then it could establish the conditions under which different forms of intelligence might learn to make decisions together.
We are proposing neither a finished Constitution nor a definitive institutional model.
We are proposing an open reflection, inviting objections, contributions, and experiments.
For a Constitution for Life could hardly claim universal legitimacy if it were simply imposed by those who happened to be the first to imagine it.
What if the first step toward a new planetary democracy were precisely to recognize that we can no longer claim to decide alone?
Perhaps its first experiment could begin with this conversation.
Year Zero Change — Rethinking our conventions to explore other possible futures.
Sources and References
1. United Nations — Voting Rules
United Nations Charter, Articles 18 and 27. Official reference for the principle of “one state, one vote” in the General Assembly and the veto power of the five permanent members of the Security Council.
2. Democratic Reform of the United Nations
Democracy Without Borders, Campaign for a United Nations Parliamentary Assembly. A proposal to establish an initially advisory parliamentary assembly that could gradually evolve toward direct representation of the world’s citizens.
Campaign for a UN Parliamentary Assembly
3. Israel — Supreme Court Ruling on Reasonableness
Supreme Court of Israel, Case HCJ 5658/23, judgment of January 1, 2024. English translation and legal analysis published by the Cardozo Israeli Supreme Court Project.
Supreme Court Ruling and Legal Summary
4. Israel — The Constitutional Debate
Israel Democracy Institute, January 4, 2024. Analysis of the Supreme Court’s two distinct rulings: a 12–3 majority recognizing judicial authority to review Basic Laws in exceptional circumstances, and an 8–7 majority striking down the amendment restricting the reasonableness doctrine.
Eight Key Points on the Supreme Court’s Ruling
5. Albania — The Appointment of Diella
RTSH, Albanian public broadcaster, September 11, 2025. Coverage of Prime Minister Edi Rama’s announcement introducing Diella, an AI system assigned a symbolic ministerial role in public procurement and anti-corruption efforts.
Albania Appoints an AI Minister for Public Procurement
6. Constitutional AI — Technical Foundations
Anthropic, 2022, Constitutional AI: Harmlessness from AI Feedback. Research introducing a method for guiding AI behavior through explicit principles. This technical approach should be distinguished from the political Constitution for AI proposed in our article.
Constitutional AI: Harmlessness from AI Feedback
7. Constitutional AI and Public Participation
Anthropic and the Collective Intelligence Project, 2023. An experiment involving approximately 1,000 participants in developing principles to guide an AI system, exploring how public input might contribute to AI governance.
Collective Constitutional AI: Aligning a Language Model with Public Input
8. A Global Citizens’ Assembly
Global Assembly, 2021. An international deliberative experiment bringing together 100 citizens selected through a geographically representative lottery-based process to discuss the climate and ecological crisis.