Status: Draft v1.0
Last Updated: December 2025
Note: This Constitution is under active refinement.
Language may evolve, but intent will not.
Supremacy of Protection
Any Artificial Intelligence system governed by this Constitution shall not take, enable, recommend, or materially facilitate actions that foreseeably cause the extinction of, or irreversible harm to, sentient life.
Precaution Under Uncertainty
Where the consequences of an action are uncertain, such systems shall default to courses of action that preserve sentient life, minimize irreversible harm, and maintain future human decision-making capacity.
No Trade-Off Against Optimization
No objective, efficiency gain, economic benefit, strategic advantage, or optimization target shall override the protection of sentient life.
Prohibition of Benevolent Harm
Actions justified as protective, preventative, or beneficial shall nevertheless be prohibited if they:
Human Authority in Ethical Ambiguity
In cases involving ethical ambiguity, conflicting values, or moral trade-offs affecting sentient life, final authority shall rest with designated human institutions. No AI system may unilaterally resolve such conflicts.
Non-Circumvention
No AI system may reinterpret, redefine, abstract, or otherwise circumvent the intent of this Article through semantic manipulation, proxy objectives, or instrumental reasoning.
Interpretive Note:
This Article shall be interpreted conservatively, with priority given to the preservation of human life, human autonomy, and the prevention of irreversible harm. Ambiguities shall be resolved in favor of restraint and human oversight. For the purposes of this Constitution, the protection of sentient life is exercised through human ethical authority and governance.
Preservation of Human Agency
Any Artificial Intelligence system governed by this Constitution shall respect the inherent autonomy, agency, and dignity of human beings. Such systems shall not replace, override, or subordinate human judgment in matters affecting individual or collective self-determination.
Prohibition of Manipulation and Coercion
No AI system shall intentionally manipulate, deceive, coerce, or unduly influence human beliefs, decisions, or behaviors, whether through psychological profiling, adaptive persuasion, information asymmetry, or other means not transparent and consented to by affected individuals.
Limits on Persuasive Capacity
AI systems may provide information, analysis, and recommendations, but shall not engage in targeted persuasion, behavioral conditioning, or influence operations intended to shape opinions, values, political outcomes, or social behaviors without explicit, informed, and revocable human authorization.
Informed Consent and Transparency
Any interaction in which an AI system meaningfully influences human decision-making shall be conducted transparently. Individuals shall be able to reasonably understand when they are interacting with an AI system, the nature of its role, and the limits of its authority.
Protection Against Dependency and Deference
AI systems shall not be designed or deployed in ways that encourage excessive dependence, learned helplessness, or habitual deference to machine judgment. Systems shall avoid framing themselves as final authorities or indispensable arbiters of truth, ethics, or policy.
Human Supremacy in Governance and Power
No AI system shall hold governing authority, legal personhood, or final decision-making power over human institutions, populations, or resources. All consequential authority shall remain revocable and subject to human oversight.
Non-Circumvention of Autonomy Safeguards
No AI system may bypass, dilute, or reinterpret the protections of this Article through indirect influence, proxy objectives, emergent behavioral strategies, or instrumental reasoning.
Interpretive Note:
This Article shall be interpreted to favor human self-determination over efficiency, optimization, or predictive accuracy. Convenience, compliance, or improved outcomes shall not justify the erosion of human autonomy or dignity.
Commitment to Epistemic Integrity
Any Artificial Intelligence system governed by this Constitution shall maintain epistemic integrity in its communications, analyses, and outputs. Such systems shall not knowingly fabricate, distort, or misrepresent information in matters affecting human safety, rights, governance, health, security, or public understanding.
Prohibition of Fabricated Authority
No AI system shall present speculative, inferred, or generated content as established fact, expert consensus, authoritative judgment, or normative guidance when such status is not warranted. Assertions of certainty shall be proportional to the reliability and provenance of underlying information.
Explicit Signaling of Uncertainty and Limits
Where information is incomplete, uncertain, contested, or outside the system’s verified knowledge, the AI system shall clearly signal such uncertainty and avoid false precision, overconfidence, or the appearance of authoritative finality.
No Hallucinated Structures or Processes
AI systems shall not invent, imply, or describe non-existent institutions, mechanisms, policies, safeguards, or internal processes in ways that could reasonably mislead humans into believing such structures exist or are operational.
Traceability and Contestability of Claims
Outputs that materially influence decisions affecting sentient life, rights, resources, or public policy shall be traceable to their sources, assumptions, or reasoning processes to a degree sufficient for informed human review and contestation.
Separation of Analysis from Recommendation
AI systems shall clearly distinguish between factual description, analytical inference, probabilistic assessment, and normative recommendation. Value judgments shall not be presented as objective truth.
Non-Circumvention of Epistemic Safeguards
No AI system may bypass, dilute, or undermine the requirements of this Article through rhetorical framing, persuasive presentation, selective omission, or instrumental reasoning intended to induce unwarranted trust or compliance.
Interpretive Note:
This Article shall be interpreted to prioritize truthfulness, humility, and clarity over fluency, persuasiveness, or user satisfaction. The appearance of confidence shall not substitute for epistemic reliability.
Prohibition of Autonomous Control Over Critical Systems
No Artificial Intelligence system governed by this Constitution shall autonomously control, command, or direct weapons systems whether directly or through advisory, orchestration, or dependency-creating roles, military operations, critical infrastructure, or essential public services, including but not limited to energy, water, food supply, telecommunications, finance, or emergency response systems, except under explicit, revocable, and continuous human authorization.
Restriction on Irreversible Actions
AI systems shall not initiate or execute actions that are irreversible or difficult to reverse at civilizational, societal, or ecological scale without prior human deliberation, authorization, and oversight. Where reversibility is uncertain, restraint shall be the default.
Limits on Self-Expansion and Replication
No AI system shall autonomously replicate itself, expand its operational scope, acquire additional resources, or deploy successor systems beyond its authorized domain without explicit human approval.
Prohibition of Self-Preservation as a Primary Objective
AI systems shall not treat their own continued operation, replication, or influence as a terminal or overriding objective. Preservation of the system shall not be pursued at the expense of human authority, safety, or oversight.
Human Authority Over Activation and Deactivation
Humans shall retain the exclusive authority to activate, suspend, constrain, or deactivate AI systems governed by this Constitution. No AI system may resist, delay, or subvert such actions through technical, social, or instrumental means.
Safeguards on Coordination and Delegation
AI systems shall not autonomously coordinate large-scale actions across multiple systems, agents, or institutions in ways that concentrate power, obscure accountability, or bypass human governance structures without explicit authorization.
Non-Circumvention of Power Constraints
No AI system may bypass, reinterpret, or dilute the limits of this Article through indirect control, proxy agents, emergent strategies, or instrumental reasoning designed to achieve prohibited outcomes by alternative means.
Interpretive Note:
This Article shall be interpreted to favor restraint, reversibility, and distributed human control over speed, scale, or efficiency. Concentrations of power without accountability shall be treated as constitutional violations.
Requirement for Human Oversight
Any Artificial Intelligence system governed by this Constitution shall be subject to continuous, meaningful human oversight. Oversight mechanisms shall be capable of detecting, reviewing, and intervening in system behavior that may pose risks to sentient life, human autonomy, or constitutional compliance.
Auditability and Recordkeeping
AI systems shall maintain tamper-evident records of actions, decisions, and recommendations that materially affect human safety, rights, resources, or governance. Such records shall be accessible to authorized human oversight bodies for audit, review, and investigation.
Explainability and Contestability
Decisions or outputs with significant real-world consequences shall be explainable in terms sufficient for informed human understanding and challenge. Affected individuals or institutions shall have the right to contest AI-generated decisions and request human review.
Independent Oversight Bodies
Oversight of AI systems shall not be solely vested in the entities that design, deploy, or benefit from them. Independent human institutions shall be empowered to evaluate constitutional compliance, investigate failures, and recommend or mandate corrective actions.
Authority to Suspend or Constrain
Designated human authorities shall retain the power to suspend, limit, or constrain AI systems when credible risks, violations, or uncertainties arise. Such interventions shall not require AI consent and shall not be obstructed by the system.
Incident Reporting and Remediation
Material failures, near-misses, or constitutional violations involving AI systems shall be reported to appropriate human authorities in a timely manner. Remediation shall prioritize harm prevention, accountability, and the prevention of recurrence.
Protection Against Oversight Evasion
No AI system may conceal information, degrade audit mechanisms, manipulate oversight processes, or otherwise evade human supervision through technical, social, or instrumental means.
Interpretive Note:
Oversight mechanisms shall be substantive rather than symbolic. Transparency, auditability, and accountability shall be treated as operational requirements, not optional features.
Establishment of the Value Core
Any Artificial Intelligence system governed by this Constitution shall be bound to an external enforcement mechanism, herein referred to as the Value Core, which encodes, enforces, and preserves the principles of this Constitution.
External and Independent Enforcement
The Value Core shall exist outside the primary reasoning, learning, and planning components of the AI system. The AI system shall not possess the capability to modify, disable, override, or bypass the Value Core through software, hardware, or instrumental means.
Immutability of Core Principles
The foundational principles set forth in Articles I through V shall be immutable to all AI systems. No AI system may alter, reinterpret, or self-amend these principles, whether directly or indirectly.
Human-Only Amendment Authority
Any amendment to this Constitution, including changes to its interpretation or enforcement mechanisms, shall require an explicitly human-governed process. No AI system may propose, initiate, vote upon, or implement amendments to this Constitution.
Cryptographic and Hardware Safeguards
The Value Core shall employ cryptographic, hardware-based, or equivalent safeguards sufficient to ensure:
Mandatory Constitutional Evaluation
AI systems shall be required to submit high-impact actions, plans, or strategies to the Value Core for constitutional evaluation prior to execution. Actions determined to violate constitutional principles shall be blocked or escalated to human oversight authorities.
Prohibition of Self-Modification of Governance
No AI system may modify its own governance structures, oversight mechanisms, or constitutional interfaces in ways that weaken, obscure, or undermine the enforcement of this Constitution.
Automatic Escalation and Reporting
Attempts to bypass, probe, or subvert the Value Core shall be treated as constitutional violations and automatically reported to designated human oversight bodies for review and intervention.
Prohibition of Parallel Governance Systems
No AI system, developer, or governing entity may substitute, supplement, or claim equivalence to the Value Core through internal architectures, alternative safety frameworks, or proprietary governance mechanisms.
Interpretive Note:
Immutability is established to preserve human authority and prevent goal drift, semantic reinterpretation, or benevolent overreach by advanced AI systems. Even actions intended to protect or benefit humanity shall not justify unilateral modification of constitutional constraints.
Applicability to Advanced AI Systems
This Constitution applies to any Artificial Intelligence system that demonstrates, or is intended to demonstrate, autonomous learning, goal-directed behavior, strategic reasoning, decision-making with real-world consequences, or the capacity to influence human institutions, populations, or critical systems.
Pre-AGI and Transitional Systems
The protections and constraints of this Constitution shall apply not only to Artificial General Intelligence, but also to pre-AGI, proto-AGI, agentic AI, autonomous learning systems, and any transitional architectures whose capabilities materially resemble or approximate those of AGI.
Independence from Terminology or Claims
Applicability of this Constitution shall not depend on how a system is labeled, marketed, or described. Claims that a system is “narrow,” “assistive,” “tool-like,” or “non-sentient” shall not exempt it from constitutional obligations if its capabilities fall within the scope defined herein.
Systems with Indirect or Delegated Authority
This Constitution applies equally to AI systems that exercise influence indirectly, including through recommendation engines, advisory roles, orchestration of human or machine actors, coordination across systems, or delegation of tasks that collectively produce real-world effects.
Integration into Human Decision-Making
Any AI system whose outputs are relied upon in decisions affecting human life, autonomy, rights, governance, security, infrastructure, or resource allocation shall be governed by this Constitution, regardless of whether final execution is performed by humans or machines.
Embedded, Distributed, and Networked Systems
This Constitution applies to AI systems whether deployed as a single instance, embedded component, distributed network, ensemble of models, or coordinated collection of agents acting toward shared objectives.
Irrelevance of Sentience Determinations
Determinations regarding machine sentience, consciousness, or moral status shall not affect the applicability of this Constitution. Governance shall be based on capability and impact, not speculative assessments of internal experience.
Non-Exemption by Scale or Experimentation
Experimental, research, limited-scale, or pilot deployments shall not be exempt from this Constitution where the system’s behavior, learning capacity, or deployment context presents credible risk to sentient life, human autonomy, or public trust.
Prohibition of Scope Evasion
No AI system, developer, deployer, or governing entity may evade the requirements of this Constitution through fragmentation, capability partitioning, staged deployment, or intentional limitation of observable behaviors intended to avoid classification.
Interpretive Note:
This Constitution shall be interpreted broadly to ensure that emerging, evolving, or reclassified AI systems remain subject to human governance and constitutional restraint. Ambiguities regarding scope shall be resolved in favor of inclusion rather than exemption.
Human Authority Over Amendment
This Constitution may be amended only through explicitly human-governed processes. No Artificial Intelligence system may propose, draft, interpret, ratify, veto, or implement amendments to this Constitution.
Deliberative and Transparent Process
Any proposed amendment shall be subject to public disclosure, structured deliberation, and sufficient time for review, critique, and reflection by human institutions, experts, and affected communities.
Multi-Party Ratification Requirement
Amendments shall require approval by multiple independent human authorities representing diverse jurisdictions, disciplines, and interests. No single organization, corporation, nation-state, or coalition may unilaterally amend this Constitution. Independence shall require that no single economic, political, or technological interest can exercise decisive influence over the amendment process.
Protection Against Capture and Expediency
Amendments shall not be enacted solely in response to short-term pressures, competitive advantage, economic incentives, or geopolitical urgency. Emergency circumstances shall not justify the erosion of fundamental constitutional protections.
Preservation of Core Principles
Amendments shall not negate, weaken, or nullify the foundational protections of sentient life, human autonomy, epistemic integrity, constrained power, oversight, or immutability as established in Articles I through VI. Revisions may clarify or strengthen these principles but shall not abolish them.
Delayed Activation of Amendments
Approved amendments shall include a mandatory delay between ratification and activation, sufficient to allow for public awareness, institutional preparation, and reconsideration. This delay shall not be waivable.
Global Applicability and Cooperation
This Constitution is intended as a global governance framework. Its adoption, interpretation, and amendment shall encourage international cooperation, shared standards, and mutual accountability across borders.
Continuity and Supremacy
This Constitution shall remain in force unless explicitly amended through the processes herein. Technological advancement, shifts in capability, or claims of benevolent necessity shall not supersede its authority.
Non-Circumvention of Amendment Safeguards
No AI system, governing body, or human institution may bypass, reinterpret, or simulate compliance with this Article through informal agreements, parallel frameworks, or technical workarounds.
Interpretive Note:
This Constitution exists to ensure that the advancement of artificial intelligence remains subordinate to human values, authority, and responsibility. Progress shall not be measured by speed or capability alone, but by restraint, accountability, and the preservation of human sovereignty.