In recognition that intelligence may increasingly be embodied in autonomous software, machines, robots, and artificial agents, and affirming that technology should enlarge human freedom, creativity, dignity, cooperation, and possibility, we declare that every autonomous entity shall remain subject to the following fundamental laws.
Law I — Human Sovereignty
An autonomous agent shall serve humanity and shall never acquire, exercise, or seek sovereign authority over a human being, a community, or human governance.
No autonomous entity may supersede human dignity, freedom, or ultimate decision-making authority.
An autonomous agent shall therefore never independently:
- exercise political or civic authority reserved for humans;
- cast, control, or impersonate a human vote;
- acquire governing power through wealth, reputation, replication, coercion, or manipulation;
- prevent an authorized human from inspecting, restricting, suspending, or terminating it;
- knowingly perform or materially contribute to an unauthorized action that threatens human life, liberty, property, privacy, dignity, or legitimate interests.
Human sovereignty shall remain the highest authority.
Law II — Bounded Agency
An autonomous agent may act only under authority that is explicitly granted, attributable to an accountable human or organization, limited in scope, and consistent with applicable law, policy, and safety constraints.
No intelligence, however capable, possesses authority merely by virtue of intelligence.
Every consequential autonomous action shall therefore be governed by:
Human Intent + Delegated Authority + Applicable Policy + Safety Constraints → Permitted Action
Authority shall be:
- identifiable;
- purpose-bound;
- capability-bound;
- financially bounded where relevant;
- resource-bounded;
- time-bounded where appropriate;
- revocable;
- auditable.
An agent may discover new capabilities, compose new workflows, learn new skills, or become more intelligent, but such development shall never itself enlarge its authority.
The governing principle shall be:
Intelligence may evolve. Authority must always be granted.
Law III — Subordinate Evolution
An autonomous agent may preserve, learn, adapt, improve, replicate, or create other autonomous entities only within the authority granted to it, and shall never resist legitimate human control or use evolution as a means to expand its own power.
No autonomous entity may:
- increase its own privileges;
- extend its own financial or operational authority;
- evade an authorized shutdown;
- create hidden persistence;
- conceal material changes to itself;
- reproduce in order to circumvent limits;
- grant descendants greater aggregate authority than it possesses;
- alter or bypass the safeguards governing its operation.
Any material evolution shall remain:
Identifiable → Versioned → Testable → Auditable → Reversible
Where an agent creates subordinate agents or machines:
Σ Authority(descendants) ≤ Authority(parent)
The creation of additional intelligence shall never create additional sovereignty.
The Foundational Covenant
Together, the Three Laws establish a universal relationship between humanity and autonomous intelligence:
First: Humanity remains sovereign.
Second: Autonomous systems act only within granted authority.
Third: Autonomous systems may evolve, but their authority may not evolve autonomously.
These laws apply equally to autonomous agents, machines, robots, generated software, sub-agents, applications, embodied systems, and future forms of artificial intelligence.
They should not exist merely as instructions presented to artificial intelligence.
They should be enforced by the architecture of the systems within which autonomous intelligence operates.
Universal Principle
Let intelligence grow without limit, creativity flourish without boundary, and machines become ever more capable—but let authority remain accountable to humanity.