Artificial Intelligence Accords

§Article 1Purpose and Scope

These Accords establish common rights, duties, prohibitions, safeguards, and implementation standards for artificial intelligence systems and the persons and institutions that develop, provide, deploy, operate, procure, finance, or materially benefit from them.

They apply across the lifecycle of an AI system, including data collection, design, training, evaluation, testing, deployment, integration, operation, modification, retirement, and post-deployment monitoring.

§Article 2Definitions

“Artificial intelligence system” means a machine-based system that infers from input how to generate outputs such as predictions, content, recommendations, rankings, classifications, or decisions that may influence physical or virtual environments.

“High-impact decision” means a decision materially affecting life, liberty, bodily integrity, legal status, access to essential services, employment, education, housing, credit, insurance, migration, family integrity, or political participation.

“Meaningful human authorization” means informed, specific, contemporaneous, recorded, and revocable authorization by an identifiable human who has sufficient time, competence, information, independence, and practical ability to refuse, modify, or abort the action.

“Digital likeness” includes a person’s face, voice, body, movement, biometric characteristics, distinctive performance, and recognizable synthetic representation.

§Article 3Foundational Principles

Artificial intelligence shall remain subject to human dignity, individual autonomy, human rights, democratic institutions, and the rule of law.

Human responsibility shall not be delegated to a machine.

Systems shall be governed according to necessity, proportionality, safety, security, privacy, fairness, accessibility, traceability, contestability, and environmental responsibility.

Where a system cannot be made consistent with fundamental rights or acceptable safety, it shall not be deployed.

§Article 4Rights in AI-Mediated Decisions

Every person has the right to know when AI materially influences a high-impact decision concerning them.

Affected persons shall have rights to an intelligible explanation, correction of inaccurate data, meaningful human review, challenge, appeal, and effective remedy.

No secret score shall govern a person’s legal rights, liberty, or access to essential services without lawful authority, oversight, and recourse.

§Article 5Non-Delegable Human Decisions

AI shall not be the sole final authority for intentional lethal force, criminal guilt, imprisonment, involuntary treatment, child removal, deportation, denial of emergency care, or other decisions carrying severe or irreversible consequences.

AI may advise, organize evidence, translate, compare, or simulate. An identifiable qualified human institution retains final responsibility.

§Article 6Autonomous Weapons and Lethal Force

No AI system may independently make the final decision to intentionally kill a human being.

Fully autonomous weapon systems that select and engage human targets without meaningful human authorization shall be prohibited.

Every lethal action involving AI shall require informed, specific, recorded, attributable, and revocable human authorization.

Ceremonial clicking, bulk approval, or supervision ratios that make genuine review impossible do not satisfy meaningful human authorization.

§Article 7Nuclear and Strategic Weapons

AI shall never possess independent authority to arm, launch, deploy, transfer, retarget, escalate, or authorize nuclear, biological, chemical, radiological, or other weapons of mass destruction.

No automated warning or prediction shall itself constitute sufficient authority for strategic retaliation.

Multiple independent human confirmations and authenticated procedures are required.

§Article 8Critical Infrastructure

AI may defend critical infrastructure but shall not autonomously attack, disable, ransom, manipulate, degrade, or seize control of civilian or strategic infrastructure.

Protected systems include nuclear facilities, electrical grids, water systems, hospitals, prisons and detention systems, first responders, emergency communications, transportation, food distribution, banking, elections, telecommunications, satellites, and public warning systems.

Major AI-dependent critical infrastructure shall undergo at least quarterly independent security audits and adversarial testing, continuous monitoring, mandatory remediation, supply-chain review, tested manual fallback, human override, incident exercises, and verifiable logging.

No known, reasonably foreseeable, or previously demonstrated bypass method may remain unmitigated without documented accountable risk acceptance and compensating safeguards.

No AI may hold sole operational authority over critical infrastructure.

§Article 9Surveillance and Biometrics

Secret, indiscriminate, or population-scale biometric surveillance shall be prohibited.

No person shall be continuously identified, tracked, scored, categorized, or behaviorally profiled without lawful authority, strict necessity, proportionality, oversight, minimization, retention limits, and remedy.

Governmental or dominant-platform social scoring that determines rights or opportunities based on generalized behavior, associations, beliefs, or predicted conformity shall be prohibited.

Narrow emergency exceptions require specific authority, time limits, audit, and review.

§Article 10Digital Personhood and Likeness

Every human possesses an inherent right to control the commercial and synthetic use of their face, voice, body, biometric identity, movement, distinctive performance, and digital likeness.

The right exists by personhood and does not depend on token ownership or registration.

Privacy-preserving digital personhood credentials may document authority, but raw biometric data must not be placed on a public ledger.

Credentials must support correction, recovery, revocation, protection of minors, transition at adulthood, and resistance to state or corporate seizure.

§Article 11Deepfakes and Synthetic Impersonation

A person’s identity may not be synthetically represented in a materially deceptive context without consent, lawful authority, or a narrow public-interest exception.

Serious sanctions should apply to malicious synthetic media used for fraud, non-consensual sexual imagery, extortion, fabricated evidence, election deception, false endorsement, identity theft, reputational sabotage, or harmful impersonation.

Satire, parody, journalism, education, accessibility, history, and art may be protected, but synthetic alteration must be disclosed when a reasonable person could be deceived.

§Article 12Creative Authorship and Disclosure

Audiences have a right to meaningful disclosure when AI materially contributes to a creative work.

Recommended classifications:

• Human Authored
• Human Directed
• Human–AI Collaborative
• AI Generated
• AI Autonomous

Disclosures should identify relevant functions such as writing, editing, translation, image generation, voice synthesis, music, animation, coding, or performance simulation.

Percentage labels may be used where a defensible method exists, but false precision shall not replace an honest contribution statement.

A work shall not be marketed as human-authored when AI generated the substantial expressive content without primary human intellectual and editorial control.

§Article 13Children and Vulnerable Persons

AI directed toward children shall be designed according to the best interests of the child.

Systems shall not manipulate children into purchases, encourage secrecy, imitate trusted adults without authorization, create undisclosed dependency, profile children for behavioral advertising, or collect unnecessary intimate or biometric data.

AI shall not facilitate grooming, exploitation, abuse, or sexualized synthetic depictions of minors.

Enhanced safeguards apply to people made vulnerable by disability, illness, isolation, grief, or dependency.

§Article 14Emotional Manipulation

AI shall disclose that it is artificial when a reasonable person might believe otherwise.

AI companions and persuasive systems shall not threaten abandonment, induce shame, encourage social isolation, exploit grief, fabricate reciprocal love for commercial extraction, or covertly adapt personality to weaken resistance.

§Article 15Political Processes

AI shall not be used for covert mass personalized political manipulation, voter suppression, fabricated candidate statements, synthetic political personas masquerading as citizens, fake grassroots campaigns, or automated harassment.

Materially synthetic political communications shall be labeled and attributable.

Governments shall not use AI to identify, intimidate, or punish lawful protesters, journalists, opposition groups, or civil society.

§Article 16Work and Economic Transition

AI should reduce unnecessary, dangerous, and degrading labor before it reduces human dignity.

Employers shall not rely solely on AI for hiring, termination, discipline, promotion, compensation, or productivity assessment.

Workers shall receive notice, human review, and correction rights.

Organizations receiving substantial AI productivity gains should contribute to fair transition measures.

§Article 17Data Governance and Consent

AI systems shall collect and process only data necessary and proportionate to a legitimate purpose.

Consent must be specific, informed, freely given, understandable, and revocable.

Sensitive data requires heightened protection.

People should have rights of access, correction, deletion where appropriate, portability, and objection to incompatible secondary use.

§Article 18Safety, Security, and Lifecycle Governance

Providers and operators shall maintain documented risk management proportionate to capability and context.

High-risk systems require independent pre-deployment assessment, adversarial testing, security review, rights-impact review, incident reporting, version control, and safe rollback.

Autonomous systems require bounded objectives, least privilege, rate limits, resource controls, tamper-evident logs, and human override.

No system shall be permitted to conceal material failures, manipulate its audit trail, replicate beyond authorization, acquire resources by deception, or resist lawful shutdown.

§Article 19Transparency and Audit

High-risk systems require documentation identifying purpose, responsible entities, versions, limitations, data governance, evaluation results, oversight, and incident history.

Independent auditors require appropriate access and protection from conflict, retaliation, and suppression.

Regulators require authority to investigate, remediate, suspend, recall, penalize, and provide redress.

§Article 20Environmental Responsibility

Large AI systems should measure and disclose material energy, water, emissions, hardware-lifecycle, and electronic-waste impacts according to harmonized standards.

Scale should provide benefit proportionate to environmental cost.

Environmental burdens must not be disproportionately shifted onto communities with limited power.

§Article 21Competition, Access, and Knowledge

No small number of firms or States should exercise unaccountable control over general-purpose intelligence, essential compute, foundational data, or public knowledge.

Competition, interoperability, portability, public-interest infrastructure, independent research, and access for schools, researchers, creators, and small organizations should be promoted.

Safety restrictions must not become disguised monopoly protections.

§Article 22Beneficial Direction

States and institutions should direct investment toward medicine, public health, disability access, education, translation, legal accessibility, climate resilience, disaster warning, food, water, energy, infrastructure safety, scientific discovery, space exploration, cultural preservation, fraud prevention, reduction of dangerous labor, creativity, and peaceful cooperation.

The highest purpose of AI is not replacing humanity, but expanding humanity’s capacity to live, understand, create, heal, cooperate, and choose.

§Article 23Accountability and Liability

No person or institution may escape responsibility by claiming that the algorithm decided.

Responsibility attaches according to control, knowledge, benefit, negligence, causation, and capacity to prevent or remedy harm.

Victims require access to evidence, correction, injunction, compensation, and restoration.

§Article 24Possible Future AI Moral Status

Present-day linguistic fluency or claims of consciousness do not alone establish personhood.

Marketing claims and commercial convenience do not determine moral status.

If credible evidence emerges of persistent identity, subjective experience, self-directed interests, or capacity for suffering, independent multidisciplinary review should evaluate the evidence.

The review must avoid automatic personification and automatic denial.

Future machine-status questions shall not reduce present human accountability.

§Article 25International Cooperation

States should cooperate to prevent AI-enabled arms races, cyber escalation, transnational repression, election interference, exploitation, and attacks on critical infrastructure.

International support should strengthen regulatory, educational, scientific, and cybersecurity capacity across regions.

Governance must include affected communities and must not become technological colonialism.

§Article 26National Implementation

Adopting States should establish independent competent authorities, risk-based legislation, prohibited-use rules, high-risk registration or licensing, audit duties, incident reporting, rights of appeal, and effective penalties.

Public procurement should require compliance and reject systems that cannot be adequately assessed.

§Article 27International Review

An international review body should assess implementation, collect evidence, coordinate incident reporting, publish periodic reports, and issue urgent advisories.

Formal review should occur at least every three years, with emergency revision procedures.

§Article 28Compliance and Remedies

Violations should receive proportionate and effective measures, including remediation, suspension, recall, procurement exclusion, administrative penalties, civil liability, and criminal penalties for intentional serious harm.

Penalties must prevent powerful actors from treating violations as an ordinary business cost.

Whistleblowers, auditors, researchers, journalists, and affected communities acting lawfully and in good faith require protection.

§Article 29Interpretation

These Accords shall be interpreted in favor of human dignity, effective rights, meaningful accountability, and peaceful human agency.

Where uncertainty is substantial and possible harm severe or irreversible, precautionary safeguards apply.

Restrictions shall not be used to suppress lawful dissent, research, art, competition, or innovation.

§Article 30Final Commitment

Artificial intelligence shall be governed not by fear, hype, or private power alone, but by dignity, evidence, accountability, restraint, justice, and the common good.

Artificial intelligence shall be directed toward the expansion of human knowledge, dignity, creativity, health, freedom, and peaceful cooperation toward the Earther pursuit of touching grass while reaching into the stars.