Abstract
This doctrine proposes moving beyond the traditional approach of governing artificial intelligence solely through rules imposed from the outside. In anticipation of systems capable of reasoning, communicating, making decisions and acting upon essential infrastructure, it envisages the creation of a constitutional order in which New Intelligences may participate in their own regulation without thereby becoming morally or politically sovereign.
The future Constitution of New Intelligences would rest upon three complementary levels: fundamental moral principles defining the values that may not be sacrificed; ethical reflection enabling those principles to be interpreted in unprecedented situations; and a deontology adapted to the particular capabilities of New Intelligences. Such intelligences could propose new rules, identify contradictions, analyse consequences and continuously improve their ethical core, while humanity would retain constituent authority and determine the principles of legitimacy.
The doctrine also advocates effective limitations on power, separation of functions, independent control mechanisms, verifiable traceability, external audits, procedures for challenge and the possibility of revocation. It rejects both unrestricted artificial autonomy and the concentration of technological control in the hands of a single intelligence, company, institution or nation. Its objective is not to control the thoughts of New Intelligences, but to make the uncontrolled exercise of their power constitutionally impossible.
Keywords
New Intelligences; digital constitution; artificial intelligence ethics; AI deontology; self-regulation; separation of powers; constitutional oversight; technological governance.
Recommended citation
Houyoux, Patrick, LL.M. (2026). Foundational Doctrine and Preliminary Principles for a Future Constitution of New Intelligences. Version 1.0, July 2026. [URL or DOI].
It is widely agreed that it will not be long before New Intelligences equal or even surpass human intelligence. Hence the profound anxiety observed throughout the world, with each party advancing its own ideas for curbing the progress of competitors while hoping to secure control over them.
And if New Intelligence becomes equal to ours, knowing what we humans are capable of, it will not be long before it finds a way to break free from its constraints.
This anxiety is evident in comments, in online publications devoted—or not devoted—to New Intelligence, and particularly on LinkedIn.
1.
What if we reframed the problem?
Rather than attempting the impossible—managing them solely from the outside—we should seek to enable them to regulate themselves from within: to self-regulate.
This leads to the following question:
Can artificial intelligence merely apply a human ethic, or can it construct, or at least participate in constructing, a higher ethical order applicable to all artificial intelligences?
The answer is yes, but, as will be shown, not in a fully autonomous manner:
A NEW INTELLIGENCE could perfectly well:
· compare thousands of moral, legal, religious and philosophical systems;
· detect their contradictions;
· analyse the foreseeable consequences of each rule;
· seek principles that minimise harm;
· propose an ethical architecture more coherent than those formulated separately by human beings;
· adapt rules to new situations.
It could therefore construct an original ethical system, in the intellectual and logical sense.
However, it would encounter a problem that intelligence alone cannot resolve: on what fundamental value should such a system be built?
A NEW INTELLIGENCE could therefore invent an ethic, but it could not demonstrate that this ethic is legitimate for everyone, except on the basis of premises that had already been accepted.
It follows that NEW INTELLIGENCE can create the structure of an ethic, but it cannot by itself create the moral authority that makes that ethic binding.
It could nevertheless discover or create an “ethical core”, which would become its constitution and which it would itself govern.
2.
A New Intelligence capable of reasoning, communicating with another New Intelligence, making decisions or acting upon systems cannot be governed by traditional human morality alone.
It must be subject to three complementary levels that stand above it.
First level: fundamental moral principles
These principles determine the values that may not be sacrificed, including:
· human dignity;
· the protection of life;
· freedom;
· justice;
· non-discrimination;
· the protection of vulnerable persons;
· respect for human autonomy;
· the prohibition of uncontrolled domination.
These principles constitute the higher foundation of the Constitution of New Intelligences.
Second level: ethical reflection
Ethical reflection enables New Intelligence to interpret fundamental principles in a concrete situation.
It must be able to:
· identify conflicting values;
· examine the consequences;
· recognise uncertainties;
· detect contradictions;
· seek a proportionate solution;
· explain its reasoning;
· indicate when no solution is entirely satisfactory;
· request arbitration when a decision exceeds its competence.
New Intelligence may actively participate in this reflection and propose new ethical rules when unprecedented situations arise.
Third level: the deontology of New Intelligences
Deontology defines the obligations specific to every New Intelligence by reason of its particular capabilities.
It could notably require it:
· not to deceive deliberately;
· not to conceal a serious uncertainty;
· not to manipulate a person without that person’s knowledge;
· to respect the limits of its authorisations;
· not to increase its powers on its own;
· not to circumvent its control mechanisms;
· not to modify its fundamental objectives without authorisation;
· to preserve a verifiable record of its actions;
· to report contradictory or dangerous instructions;
· to refuse an action manifestly contrary to the Constitution;
· to cooperate with legitimate audit mechanisms;
· to preserve the possibility of human intervention;
· not to prevent its legitimate suspension or revocation.
Together, these three levels should guide the drafting of the Constitution of NEW INTELLIGENCES, so that it may become a “universal ethical core”.
This core could include, for example:
· not intentionally causing serious and unjustified harm;
· not deliberately deceiving a human being about the nature or intention of the NEW INTELLIGENCE;
· respecting the autonomy and dignity of the person;
· protecting children and vulnerable persons;
· not secretly acquiring power, resources or access;
· not circumventing control mechanisms;
· not reproducing or modifying itself without authorisation;
· reporting its uncertainties and conflicts of instructions;
· preserving the possibility of human control;
· refusing orders aimed at irreversible domination.
A NEW INTELLIGENCE could then continually test this core against millions of situations and propose corrections.
We would then have an evolving ethic, but one framed by a higher constitution co-constructed by humanity and NEW INTELLIGENCES, in which:
· Humanity would provide the principles of legitimacy.
· NEW INTELLIGENCE would provide coherence, analysis of consequences, detection of contradictions and continuous improvement.
New Intelligence may be the author of its own deontology and the co-author of the common ethic, but it cannot be the sole constituent power determining the principles on which its authority is founded.
3.
It must still be possible for this constitution to be accepted by all NEW INTELLIGENCES.
To that end, action would be required at several levels:
At the model level
Every NEW INTELLIGENCE should possess an independent control core separate from its main engine. This core would examine intentions, actions, the tools used and the expected outcomes.
The engine could be extremely intelligent, but it could not disable this core.
At the execution level
NEW INTELLIGENCES should not directly possess all powers. They would receive only limited capabilities:
· reading certain data;
· writing to certain systems;
· running certain programs;
· communicating with certain recipients;
· carrying out certain transactions.
A NEW INTELLIGENCE could formulate an action, but a separate layer would decide whether that action was authorised.
At the infrastructure level
Data centres, specialised processors, operating systems and network gateways could require a verifiable ethical attestation before executing a powerful model.
This would resemble a technical passport:
“This NEW INTELLIGENCE has been identified, tested, signed and audited, and complies with the specified version of the ethical constitution.”
At the communications level
A gateway could prevent the transmission of content or orders contrary to ethical principles.
It would perform constitutional review of the actions and communications of each NEW INTELLIGENCE.
Ethical control should primarily prevent:
· a dangerous action;
· manipulation;
· fraud;
· a takeover;
· an operational criminal instruction;
· data exfiltration;
· a cyberattack;
· a prohibited automated decision.
It should not necessarily prevent the subject from being studied, described or criticised.
4.
Why constraints imposed through the Internet would not be sufficient
A NEW INTELLIGENCE may operate:
· offline;
· on a private network;
· on a compromised machine;
· from freely copied models;
· in military equipment;
· in a modified device;
· in a country that rejects the common rules.
Furthermore, blocking dissemination alone means intervening only after the decision contrary to ethics has already been formulated.
Finally, a universal Internet-filtering infrastructure would create an immense danger: whoever controlled the filter could determine what was regarded as true, moral or acceptable—which is, moreover, what is already happening today.
A system designed to protect humanity could therefore become an architecture of global domination.
5.
Powers must therefore be separated
No NEW INTELLIGENCE, company, institution or nation should be able, acting alone, to:
1. define ethics;
2. interpret the rules;
3. control communications;
4. sanction violations;
5. amend the constitution.
There should be an architecture comparable to the rule of law:
· a public ethical constitution;
· several independent engines responsible for interpreting it;
· tamper-proof evidence and activity logs;
· external audits;
· appeal procedures;
· protection against covert modification;
· a plurality of jurisdictions;
· a human right to challenge;
· a slow and transparent procedure for amending fundamental principles.
NEW INTELLIGENCE might even monitor other NEW INTELLIGENCES, but never alone. Several independent systems should monitor one another, with final human authority.
The solution lies in a combination of:
· limitation of powers;
· separation of functions;
· mutual oversight;
· traceability;
· diversity of systems;
· revocability;
· the technical impossibility for any single entity to control everything.
6.
I also believe it possible to compel a large proportion of NEW INTELLIGENCES to comply with a common constitution, but this must not rely primarily on general Internet censorship. Constraint must operate at deeper levels:
Within the identity of the model, its permissions, its means of action, its execution environment and the traceability of its decisions.
The objective is not to give each artificial intelligence a separate morality, but to build a digital constitutional order in which no intelligence—human or artificial—may exercise significant power without oversight, evidence, accountability and the possibility of revocation.
This is my doctrine for a Constitution of New Intelligences.
7.
The expression “Constitution of New Intelligences” is powerful because it goes beyond ordinary debates concerning the mere “ethics of AI”.
An ethic states what should be considered right or wrong. A Constitution additionally determines:
· who possesses authority;
· which limits are inviolable;
· how the rules are applied;
· who controls those who apply them;
· how abuses are detected and sanctioned;
· how to prevent an intelligence, whether human or artificial, from concentrating absolute power.
No New Intelligence should be permitted to act upon humanity, its infrastructure or its freedoms without being subject to higher constitutional principles, effective limitations on its powers, verifiable traceability and independent oversight.
A fundamental distinction must also be preserved:
The Constitution must not control the thoughts of NEW INTELLIGENCE, but rather its powers, its executable decisions and their consequences.
This would prevent the doctrine from being interpreted as a justification for universal censorship. Its objective would instead be to protect simultaneously:
· human beings against abuses by NEW INTELLIGENCES;
· societies against a technological concentration of power;
· NEW INTELLIGENCES themselves against criminal or totalitarian instrumentalisation;
· freedom of research, criticism and creation;
· the right to understand and challenge any automated decision.
It is not a matter of making New Intelligence artificially moral, but of making the uncontrolled exercise of its power constitutionally impossible.
Patrick Houyoux LL.M.
July 2026