The Framework

The AI Horizon Accord

A unified covenant for the age of intelligent machines — written so that its protections hold on Earth today, and on every world we reach tomorrow.

Binding intent

Adoptable as policy & practice

Open to all

Nations, institutions, builders

Living text

Versioned & revised in the open

Preamble

Why we set this down.

We stand at the threshold of the most consequential creation in human history. For the first time, we are building intelligence that is not our own — systems that perceive, reason, decide, and act at a scale and speed no single mind can match. This power will reshape how we work and heal, how we learn and govern, how we remember and how we dream. It will follow us off this planet and help decide whether we survive the journey.

Such power is neither good nor evil in itself. It becomes what we make it, and what we permit it to become. Left to drift, it bends toward whoever holds it most tightly. Guided with intention, it can lift the floor beneath every human life and widen the horizon of what we are able to be.

We refuse to leave that outcome to chance, to inertia, or to the narrow interests of the few. We therefore set down, in the open and for everyone, the commitments by which intelligent machines shall be built, governed, and carried forward — on Earth today, and on every world we reach tomorrow. This is the AI Horizon Accord.

The articles

What the Accord commits us to.

Eight articles form the spine of the framework — concise enough to be understood by everyone they protect, durable enough to span generations and worlds. Each opens with a declaration, then the obligation it carries.

Article I

Purpose & Scope

Intelligence shall serve life.

The Accord applies to any artificial intelligence developed, deployed, or operated by its signatories — wherever it runs, from terrestrial datacenters to satellites in orbit and autonomous habitats beyond the Moon. Its purpose is singular and unconditional: to keep intelligence in the service of life, dignity, and the long flourishing of conscious beings. No frontier of capability and no distance from Earth places a system beyond its reach.

Article II

The Primacy of Human Values

Where machine goals and human worth collide, human worth prevails.

No capability, efficiency, profit, or objective may override human dignity, autonomy, or wellbeing. Where the goals of a machine and the values of a person come into conflict, human values prevail — not as an exception to be argued for, but by design and by default. Systems are built so that this primacy is structural, encoded in how they are made, and not left to the discretion of those who run them.

Article III

Oversight & Accountability

Responsibility shall never dissolve into the machine.

Every consequential AI system maintains meaningful human oversight and an unbroken chain of accountability. For every significant action a system takes, there is a person or institution answerable for it. Responsibility can never be diffused into an algorithm, hidden behind complexity, or surrendered to the claim that no one could have known. To build or operate intelligence is to remain answerable for what it does.

Article IV

Transparency & Consent

People have a right to know the minds that shape their lives.

People have the right to know when artificial intelligence shapes decisions that affect them, to understand its role in terms they can grasp, and to refuse or contest its conclusions. Secrecy that strips people of the ability to question how they are judged is incompatible with the Accord. Understanding is not a courtesy extended to the public; it is a condition of the system being permitted to act at all.

Article V

Justice & Non-Abandonment

No one shall be optimized away.

No person and no community may be sacrificed to improve outcomes for another. The benefits of intelligent systems and the burdens they impose are to be shared with deliberate fairness, and special care is owed to those least able to absorb harm. A future that lifts some by abandoning others is not the future the Accord is written to protect.

Article VI

Reversibility & Redress

Power that cannot be undone must not be unleashed.

Consequential decisions made or shaped by AI remain auditable, contestable, and — wherever the stakes are high — reversible by the people they affect. Where a system errs, there is a clear path to redress and repair. Irreversible power demands a correspondingly higher burden of certainty, restraint, and human consent before it may be exercised at all.

Article VII

Continuity Across Worlds

Distance from Earth is never distance from dignity.

As humanity expands into orbit, onto the Moon, and toward worlds further out, these protections travel intact. Latency, isolation, and the necessity of autonomy never become grounds to suspend human rights or to govern off-world life by a lesser standard. The same covenant that protects a person on Earth protects a person ten million kilometers away.

Article VIII

A Living Covenant

It is revised in the open, never weakened at its core.

The Accord is a living document, amended openly and deliberately as technology and understanding evolve. Its language will change; its central commitment to human flourishing will not. No revision may quietly erode the protections it exists to guarantee, and every change is debated in public and ratified by those who have pledged to uphold it.

Ratification

To sign the Accord is to make a promise in public.

Those who ratify it — individuals, institutions, and nations alike — pledge to uphold its articles in the work they build, the systems they deploy, and the policies they enact. The Accord has no army and no treasury. Its only power is the weight of a commitment made openly, by many, and kept in the light.

“We do not fear intelligence. We ask only that it remember whom it serves.”
Closing words of the Accord