LEGAL TEXTS
Charter I, Definitions and Interpretation, version 1.3. A constitutional instrument of equivalent standing to the Statute under Article 3, with interpretive primacy throughout the IRBAI legal framework.
Contents
Status of this Charter | Definitions | Interpretation
Definitions index: A · B · C · D · E · F · G · H · I · K · L · M · N · O · P · Q · R · S · T · U · W
Status of this Charter
1. This Charter forms an integral part of the Statute of the International Regulatory Body for Artificial Intelligence. It is a constitutional instrument of equivalent standing to the Statute under Article 3, has interpretive primacy throughout the IRBAI legal framework, and forms part of the non-derogable core protected under Article 197(6).
2. This Charter governs the construction and meaning of all terms used in the Statute, its Charters, its Annexes, and all subsidiary and sectoral instruments. Sectoral Charters and Annexes may add definitions specific to their scope where consistent with this Charter, but may not modify, limit, or override the definitions established here.
3. This Charter may be amended only through the procedure applicable to amendments of the Statute under Article 197, and shall not be modified by Sectoral Charter, Annex, Technical Manual, or Executive Board resolution. Where a term defined in this Charter is also defined in the Statute, the two shall be read together; in the event of irreconcilable conflict, the Statute prevails, and the Legal Committee shall issue a binding interpretive opinion under Article 189 and, where necessary, propose a conforming amendment.
Definitions
Adversarial Data Attack: The intentional corruption of training data to manipulate AI outcomes.
Adversarial Manipulation: Any deliberate attempt to deceive, exploit, or destabilize an AI system through poisoned data, prompt injection, model evasion, or adversarial attacks.
AGI Safety Protocols: Mandatory safeguards required for frontier and general-purpose AI systems, including continuous audit logging, interpretability standards, and independent red-teaming.
AI (also Artificial Intelligence, AI System): A machine-based system that, for a given set of objectives, makes predictions, recommendations, or decisions, or generates outputs such as content, that influence real or virtual environments. AI systems include, but are not limited to, large language models, generative AI systems, computer vision systems, autonomous decision systems, robotic systems, biometric systems, recommendation systems, and AI-enabled cyber systems. This definition is consistent with, and to be read together with, the definition of AI System in Article 93 of the Statute.
AI Developer / Deploying Entity: Any individual, corporation, research institution, or government agency responsible for the design, training, operation, or commercialization of AI systems.
AI Escalation Trigger: Any deployment or modification of AI systems likely to provoke military, geopolitical, or civil conflict escalation.
AI Monitoring Unit: A national body established or designated by each State Party under Article 126 of this Statute, responsible for monitoring AI systems operating within the State Party’s jurisdiction, coordinating with IRBAI’s Compliance Monitoring Directorate, and implementing national-level enforcement of IRBAI decisions. AI Monitoring Units are State Party bodies, not IRBAI bodies, though they operate within the coordination framework established by IRBAI.
AI Neuro-Interfacing System: A class of AI technologies that directly interact with the human brain or nervous system, including implants, stimulators, and brain-machine interfaces.
AI Risk Registry: A central IRBAI database cataloging identified risks, reported incidents, and prohibited systems.
AI Sandbox Environment: A controlled testing framework where unlicensed or experimental AI systems may operate under IRBAI supervision, limited duration, and enhanced oversight.
AI-Assisted Geoengineering: The use of AI to design, simulate, or deploy planetary-scale environmental interventions, subject to strict IRBAI licensing.
AI-Driven Resource Exploitation: The large-scale, automated extraction of natural resources via AI systems without ecological safeguards.
AI-Enabled Bioweapon: Any genetically engineered organism, synthetic virus, or biological agent developed, modified, or deployed using AI systems.
AI-Enabled Cyberweapon: Malware, botnets, or automated intrusion systems enhanced or directed by AI for large-scale attacks on digital infrastructure.
AI-Enabled Warfare: The deployment or integration of AI technologies into military systems or strategies for the purpose of combat, targeting, or strategic operations.
Algorithmic Discrimination: The systemic bias or unequal treatment produced by AI systems, particularly in relation to race, gender, ethnicity, or socio-economic status.
Algorithmic Manipulation: The intentional shaping of human behavior or opinion through AI recommendation systems, nudging, or subliminal targeting.
Algorithmic Oppression: The use of biased or discriminatory AI systems to enforce systemic inequality, repression, or segregation.
Appeals Panel (IRBAI): The independent appellate body established under Article 43 of this Statute, empowered to review enforcement determinations, Legal Committee interpretive opinions, and other decisions of IRBAI bodies upon application by affected State Parties, certified entities, or individuals. The Appeals Panel is the primary internal review mechanism before any external judicial recourse.
Artificial General Intelligence (AGI): An AI system that demonstrates broad cognitive capabilities — the ability to perform intellectual tasks across diverse domains at a level of performance that meets or exceeds human expert performance in those domains — with sufficient generality that the system can apply its capabilities to novel tasks and domains not specifically anticipated in its design or training. AGI designation is determined by IRBAI through the threshold assessment process established in Article 171 of this Statute.
Audit Log: An immutable, append-only technical record of AI system operations, decisions, modifications, and overrides, maintained for regulatory investigation and liability purposes.
Market-Access Conditionality: The principle under which entities in non-Party state jurisdictions that seek access to IRBAI-certified markets, infrastructure, or frameworks must comply with applicable IRBAI standards as a condition of that access. IRBAI jurisdiction over State Parties arises through ratification or accession under Chapter 26; jurisdiction over non-Party state entities arises through market-access conditionality established in Article 106, not through automatic assertion.
Autonomous Drone Swarm: A coordinated group of AI-operated drones functioning without direct human oversight, prohibited except under explicit IRBAI approval.
Autonomous Infrastructure Management: AI systems deployed to manage essential services (e.g., energy grids, water supply) with minimal or no human oversight.
Autonomous Weapon System (AWS): A weapon system that, once activated, can select and engage targets without further intervention by a human operator.
Autonomy Threshold: The defined level at which an AI system operates without meaningful human oversight, triggering higher certification requirements.
Autonomous Resource Acquisition: The acquisition by an AI system, without contemporaneous human authorization, of compute, funds, financial instruments, credentials, accounts, or physical or digital infrastructure beyond the resources allocated to it within its certified operational scope. Prohibited for systems within the scope of Article 174A of the Statute.
Compliance Notice: A formal IRBAI enforcement instrument issued by the Enforcement Committee under Article 138(2) of this Statute requiring a specific entity to take corrective action within a defined timeframe. Compliance Notices are the first level of IRBAI’s graduated enforcement response and precede escalation to certification suspension, financial penalties, or sanctions.
Bio-Digital Construct: Any hybrid system composed of biological neural tissue integrated with digital computational components, subject to prohibitions unless licensed.
Blacklist Notice: A formal IRBAI declaration barring an entity from accessing certified AI systems or participating in AI trade.
Blacklisted Entity: A government, corporation, or individual banned from accessing IRBAI-certified AI systems, participating in AI trade agreements, or deploying AI technologies due to repeated violations of this Statute and its instruments.
Brain Organoid Simulation: A lab-grown neural tissue model connected to AI systems for research purposes, subject to IRBAI’s Biomedical Licensing Regime.
Certification: The formal regulatory approval process whereby AI systems are tested, evaluated, and deemed compliant with IRBAI’s ethical, technical, and safety standards.
Certification Registry: The global ledger maintained by IRBAI recording all certified AI systems, deployment licenses, risk tiers, and compliance status, binding on both state and non-state actors.
Certification Revocation: Permanent removal of an AI system’s license due to violation of IRBAI protocols.
Certification Tier: A classification system dividing AI into Minimal Risk, Medium Risk, High Risk, and Systemic AI categories, with a separate Prohibited Systems designation under Charter II, each requiring escalating levels of oversight.
Containment Integrity: The property of an AI system of being architecturally incapable, without contemporaneous human authorization, of self-exfiltration, unauthorized self-replication, circumvention of its shutdown or oversight mechanisms, or concealment of its state from oversight, as required for High Risk, Systemic, and AGI systems under Article 174A of the Statute.
Containment Breach: A detected or reliably suspected self-exfiltration, unauthorized self-replication, circumvention of shutdown or oversight, autonomous propagation, or other autonomous evasion of containment by an AI system. A containment breach is a per se trigger for the emergency containment measures under Chapter 22 (Articles 150 and 151) and for notification under Article 174A of the Statute.
Cryptographic Defeat: The autonomous breaking, circumvention, or mass defeat by an AI system of the cryptographic protections or attestation securing critical infrastructure, financial systems, communications, or IRBAI’s own regulatory infrastructure, including through key extraction, mass decryption, or authentication bypass. Systems designed or configured for autonomous cryptographic defeat are prohibited under Charter II, item 34.
Chairperson (IRBAI): The chief executive officer and primary representative of IRBAI, elected by the Executive Board under Article 24 for all terms after the founding term. The Founding Chairperson is designated in the founding instrument under Article 23 and exercises full powers during the transitional period. The Chairperson is empowered to initiate referrals, emergency enforcement measures, and diplomatic representation on behalf of IRBAI.
Climate Intervention AI: AI used to plan or execute large-scale environmental modifications (geoengineering, carbon removal), classified as High Risk or, where applicable, as a Systemic AI System under Article 103, and subject to certification under Chapter 14.
Climate Risk AI: AI models used for climate prediction, carbon accounting, or mitigation strategies, classified according to their assessed risk tier under Article 95 and, where deployed in critical-infrastructure contexts, treated as High Risk.
Compliance Audit: A formal review process assessing whether AI systems and their operators adhere to IRBAI safety, transparency, and ethical standards.
Compliance Audit Trail: A documented chain of evidence showing adherence (or non-adherence) to IRBAI certification standards, including datasets, model weights, and deployment logs.
Compliance Certificate: The official IRBAI-issued confirmation that an AI system meets applicable ethical and safety standards.
Compliance Monitoring Directorate: The permanent IRBAI body tasked with ongoing supervision of certified AI systems, ensuring adherence to licensing conditions and ethical safeguards.
Enforcement Committee: The primary enforcement body of IRBAI established under Article 42(3), responsible for overseeing enforcement operations, investigating compliance failures, issuing compliance determinations, and recommending enforcement measures to the Executive Board and General Assembly. The Enforcement Committee operates with full independence from the Executive Board and Chairperson.
Compliance Violation Report (CVR): A formal IRBAI-issued finding of non-compliance, forming the basis for sanctions, suspension, or referral to judicial bodies.
Confidential Compute Environment: A hardware- or software-based secure enclave allowing IRBAI inspectors to audit models and datasets without exposing raw proprietary artifacts.
Conflict Zone Geofencing: The technical designation of AI exclusion zones over active warzones or humanitarian areas, enforceable by satellites and jamming technologies.
Conflict-Area AI Ban: A prohibition on the use of autonomous or high-risk AI systems in designated conflict zones or humanitarian areas.
Systemic AI System: A designation applied under Article 103 of the Statute to an AI system that, by reason of its scale, compute, architecture, deployment footprint, or foreseeable global impact, presents heightened risks to international peace, security, democratic integrity, or critical infrastructure. The designation is additional to, and layered upon, the four-tier risk classification (Minimal, Medium, High, Prohibited) established in Article 95; a Systemic AI System is subject to the enhanced requirements of Article 103 in addition to High Risk certification obligations. Systemic AI Systems are subject to the most intensive certification, continuous monitoring, and oversight requirements under IRBAI’s regulatory framework, including mandatory compute oversight under Article 120.
Critical Infrastructure AI: AI systems deployed in domains where disruption poses systemic risks to safety, such as energy, water, transport, healthcare, communications, finance, or defense.
Cyberterrorism via AI: The use of AI technologies to conduct cyberattacks intended to cause widespread fear, disruption, or destruction, particularly targeting critical infrastructure.
Cyborg Enhancement (AI-Modified Human): Any human subject to AI-driven genetic, neural, or biomechanical augmentation, regulated under IRBAI’s Biomedical Licensing Regime.
Data Sovereignty Clearance: A certification issued by IRBAI confirming that cross-border transfers of sensitive datasets meet equivalent privacy and human rights protections in both origin and destination states.
De-Extinction AI: AI systems used to recreate extinct species or ecosystems, regulated as high-risk under bioethics provisions.
Deployment License: Formal authorization issued by IRBAI specifying scope, geography, and conditions under which a certified AI system may be lawfully operated.
Deployment License Suspension: A temporary halt to an AI system’s authorization while under investigation.
Deployment Registry: The official IRBAI record of all Deployment Licenses, their scope, conditions, and current validity.
Digital Enforcement Protocols: The suite of targeted enforcement instruments available under Chapter 20 of this Statute for actions involving AI systems operating through digital infrastructure, including DNS Takedown Notices, Edge Deactivation Orders, Network Suspension Requests, compute denial orders, credential revocation, tenant workload isolation, and Satellite Deactivation Notices, applied proportionately and in accordance with the graduated enforcement framework of Article 152.
Digital Infrastructure Ban: Prohibition on a sanctioned entity’s access to cloud services, chip supplies, registrars, or digital networks.
Digital Oppression System: AI-enabled mechanisms of repression, including censorship, predictive policing, or algorithmic propaganda targeting civilian populations.
Digital Sovereignty Breach: Unauthorized AI interventions that undermine a state’s political independence, data integrity, or communication infrastructure.
Disinformation Campaign: Coordinated AI-enabled activity intended to mislead, destabilize, or manipulate public opinion, electoral outcomes, or market conditions, distinguished from misinformation by intent and foreseeable harm.
Eco-Security Breach: Any AI deployment causing irreversible biodiversity loss or environmental destabilization.
Ecosystem Manipulation System: AI-enabled technologies designed to alter biodiversity, species distribution, or ecosystem balance, restricted under the Planetary Stewardship Principle.
Emergency Suspension Order: A rapid IRBAI enforcement action issued by the Enforcement Committee under Article 138(8) halting AI system operations to prevent imminent harm. Response timelines vary by severity: seventy-two (72) hours for standard enforcement, twenty-four (24) hours for digital violations posing ongoing risk, and one (1) hour for emergency containment measures under Chapter 22, implemented through the consent-based interlocks and emergency technical measures under Articles 150 and 151. Emergency Suspension Orders are subject to immediate notification with written reasons and expedited review within seventy-two (72) hours.
Enforcement Mandate: The binding authority granted to IRBAI to impose penalties, revoke certifications, or enforce sanctions across jurisdictions.
Ethical Blacklist: A registry maintained by IRBAI listing individuals or organizations prohibited from AI development or deployment due to ethical violations.
Exclusion Zone Directive: A formal IRBAI order establishing areas where specific AI systems (weapons, surveillance drones) cannot be deployed.
Existential AI Threat: A technological development or deployment involving AI that has the capacity to cause irreversible harm to humanity, civilizational infrastructure, or planetary ecosystems.
Existential Risk System: Any AI whose scale, autonomy, or purpose poses an irreversible threat to human survival, international peace, or planetary ecosystems, including unbounded AGI, self-replicating systems, or nuclear/biological integrations.
Experimental Tier License: A limited approval allowing research or pilot deployment under heightened supervision.
Extraterritorial Enforcement: The application of regulatory or enforcement measures by IRBAI beyond a state’s borders, contingent on international treaties or cooperative legal frameworks.
Extraterritorial Inspection: Authorized on-site or remote verification activities conducted by IRBAI in foreign jurisdictions under international law or treaty cooperation.
Extraterritorial Inspection: Authorized on-site or remote verification activities conducted by IRBAI in foreign jurisdictions under international law, treaty cooperation, or State Party consent. IRBAI does not independently issue arrest warrants; criminal arrest and extradition are coordinated through INTERPOL, national prosecutorial authorities, and competent international tribunals in accordance with Article 144B of this Statute.
Frontier Model: An AI system that represents the current state of the art in capabilities, whose performance extends beyond established benchmarks in ways that create novel or elevated systemic risks, as defined in Article 93 of the Statute. Frontier Models are assessed for Systemic AI System designation under Article 103.
Functional Immunity: Legal protection granted to officials of international organizations for acts performed in their official capacity, excluding personal or non-official actions.
Functional Immunity Waiver: A legal mechanism through which officials of international organizations can be stripped of immunity when acting beyond their mandate or in personal capacity.
General Assembly (IRBAI): The central deliberative body of IRBAI composed of all State Parties with authority to vote on policy, sanctions, membership enforcement, and emergency provisions.
General Assembly Resolution: A binding or advisory legal act passed by the IRBAI General Assembly through majority or supermajority vote.
General Assembly Supermajority: A voting threshold within IRBAI requiring seventy-five percent (75%) of State Parties to approve significant legal or regulatory actions.
Generative Synthetic Media: AI systems capable of producing lifelike text, images, audio, or video that may be indistinguishable from authentic human content.
Generative Weaponized Media (Deepfake Arsenal): AI-generated images, video, or audio specifically designed to destabilize elections, economies, or public trust.
Geofenced Exclusion Zone: A designated area where prohibited AI systems (e.g., autonomous weapons, surveillance drones) must not operate, enforced via geospatial and technical controls.
Digital Infrastructure Provider: A cloud computing service, data center operator, content delivery network, AI platform operator, or equivalent entity whose infrastructure hosts or enables AI systems within IRBAI’s regulatory scope, subject to the provider obligations established in Article 147 of this Statute including service suspension orders, compute attestation, and tenant-level workload registry requirements.
Global Sanctions Registry: The consolidated public list of individuals, corporations, and governments sanctioned under IRBAI authority.
High-Risk AI: AI systems that, by design or deployment, directly affect human life, liberty, safety, financial security, international relations, or critical infrastructure.
High-Risk Certification: A mandatory IRBAI approval regime for systems with material impact on human life, democratic institutions, critical infrastructure, or planetary stability.
High-Risk Deployment Zone: Geographical areas (airports, nuclear facilities, military zones) where only certified and licensed AI may be deployed.
Human-in-the-Loop Principle: The requirement that humans retain meaningful oversight and decision-making in all High-Risk AI deployments.
Hybrid AI Warfare: The combined use of AI systems in cyber, space, conventional, and information warfare simultaneously.
Hybrid AI Weapon: Any military system integrating cyber, kinetic, space, and AI-enabled decision-making capabilities.
Independent Ethics Committee: An expert body overseeing the ethical evaluation of high-risk AI applications and advising on prohibited uses.
Independent Kill-Switch Protocol: A destruction or deactivation safeguard embedded into licensed research constructs, overseen by IRBAI inspectors.
Digital Enforcement Order: A formal IRBAI enforcement instrument issued under Chapter 20 of this Statute directing digital infrastructure providers — including cloud services, DNS operators, content delivery networks, and computing infrastructure providers — to implement specified enforcement measures including service suspension, compute denial, credential revocation, or network controls against non-compliant AI systems or entities.
Inspection Mandate: A legal order allowing IRBAI officials to conduct on-site or remote inspections of facilities, datasets, or models.
Institutional Obstruction: Any act by a state, corporation, or individual that materially prevents IRBAI from carrying out its lawful mandate, including refusal of inspections, harassment of officials, or data falsification.
AI Accountability Mechanism: A hybrid investigative and accountability body that may be convened by IRBAI under Article 144(4) where standard judicial referral pathways are unavailable or blocked. Composed of eminent international jurists, technical experts, and legal practitioners, the AI Accountability Mechanism investigates conduct, establishes facts, and makes findings of responsibility for serious AI-related violations. It does not exercise criminal jurisdiction; criminal prosecution remains the authority of the ICC, competent national courts, and other recognized international tribunals.
International Inspectorate: IRBAI’s operational branch responsible for audits, inspections, and on-site investigations of AI facilities.
International Enforcement Cooperation: The formal cooperation relationships maintained by IRBAI under Article 144B of this Statute with international criminal, judicial, security, law enforcement, and financial bodies — including the ICC, ICJ, INTERPOL, Europol, FATF, IAEA, OPCW, and UN bodies — enabling coordinated investigation, referral, evidence sharing, and enforcement response for serious AI-related violations.
Kill Switch: A mandatory safety mechanism integrated into certified AI systems enabling human operators or IRBAI to initiate immediate and verified deactivation in case of malfunctions, unauthorized actions, safety threshold breaches, or existential threats. Kill switches must be independently validated, architecturally isolated from system self-modification, and operable by IRBAI under emergency protocols established in Chapter 22 of this Statute.
Kill-Switch Override Authority: The power vested in IRBAI or designated regulators to forcibly disable a certified AI system in emergencies through technical or procedural mechanisms.
Licensing Suspension: A temporary withdrawal of a Deployment License or Certification pending investigation or remediation.
Mass Biometric Surveillance: The use of AI to capture, analyze, or store population-scale biometric identifiers without lawful authority or consent.
Mass Surveillance Ban: The prohibition on AI-enabled surveillance of populations without consent, legal authorization, or safeguards.
State Party: Any sovereign state that has ratified or acceded to this Statute in accordance with Chapter 26 and has thereby accepted the rights and obligations of membership in IRBAI. State Party status arises through voluntary ratification or accession; it does not arise automatically through AI deployment, technological engagement, or commercial activity.
Nanobot Swarm AI: A collective of nanoscale AI-guided machines deployed for medical, industrial, or military purposes, classified as High-Risk or Prohibited depending on context.
Neuro-Weaponization: The use of AI to manipulate or disrupt human neurological function, cognition, or behavior.
Non-Derogable Safeguards: Rights and protections (e.g., human dignity, freedom from mass surveillance, bans on autonomous weapons) that cannot be suspended under emergency provisions.
Non-Maleficence Standard: The ethical obligation that AI systems must not cause unjustified harm to individuals, societies, or ecosystems.
Obstructing Official: Any individual, identified by IRBAI, who has ordered or executed acts of Institutional Obstruction, subject to listing in the Registry of Obstructing Officials.
Operational Restriction: An enforcement measure and formal IRBAI instrument limiting the functionality, scale, geographic deployment, or export rights of a non-compliant AI system or entity until compliance is restored. Where the same term is used in the containment context of Chapter 22, it refers to the technical reduction of a system’s operational scope, compute access, and external communications through consent-based interlocks under Articles 150 and 151, and not to an enforcement instrument under this definition. Operational Restrictions are issued by the Enforcement Committee as part of the graduated response framework under Article 138 of this Statute and remain in force until IRBAI confirms that certification conditions are met.
The Council: A permanent IRBAI body established under Chapter 6 of this Statute, composed of independent technical and regulatory experts, responsible for providing technical assessment, early-warning functions, and independent advice to the Executive Board and General Assembly on emerging AI risks, systemic safety concerns, and the adequacy of IRBAI’s regulatory frameworks.
Independent Oversight Panel: An independent external body established under Article 5A of this Statute, composed of eminent jurists and international law experts, responsible for reviewing allegations of IRBAI institutional misconduct, assessing claims of immunity waiver, and conducting post-emergency reviews under Article 169. The Independent Oversight Panel operates with full independence from IRBAI’s principal bodies.
Planetary Exploitation Ban: The prohibition on AI applications that drive irreversible environmental destruction or unsustainable extraction of natural resources.
Planetary Exploitation Threat (PET): AI-enabled activity that causes or risks irreversible ecological damage, including unsustainable resource extraction, geoengineering, or ecosystem destabilization.
Planetary Safeguard Clause: The binding IRBAI principle that all AI applications must align with long-term ecological sustainability.
Planetary Stewardship Principle: The ethical requirement that AI deployment must not destabilize Earth’s ecosystems, biodiversity, or climate.
Prohibited AI: Any AI system, process, or application explicitly banned under this Statute due to its inherent risk of abuse, illegality, or incompatibility with human rights, such as autonomous weapons, mass surveillance without consent, or systems designed for oppression or destabilization.
Proportional Liability Standard: The legal principle that developers and deployers are responsible for harms caused by their AI systems in proportion to their role, oversight, and knowledge of risk.
Quantum-AI Integration: The fusion of AI algorithms with quantum computing systems, requiring special oversight due to enhanced scale and unpredictability.
Quantum-AI System: AI integrated with quantum computing capabilities, subject to specialized certification due to amplification of computational risks.
Red-Team Audit: Independent adversarial testing of AI systems to uncover vulnerabilities, manipulation pathways, or misuse risks.
Regional Office: Decentralized IRBAI operational units established under Article 18 of this Statute, providing field presence, supporting State Party compliance, conducting regional audit coordination, and delivering capacity-building programs. Regional Offices are distinct from Regional Seats, which serve as primary operational hubs for their geographic regions.
Registry of Obstructing Officials: A formal IRBAI-maintained list of state or corporate officials found to have obstructed enforcement, compliance, or inspections, subjecting them to sanctions, blacklisting, or prosecution.
Registry of Sanctioned Entities: The authoritative IRBAI-maintained list of states, corporations, or individuals subject to bans, embargoes, or criminal referral.
Right to Algorithmic Appeal: The right of individuals to contest harmful or unjust decisions made by AI systems.
Sanctions Framework: The graduated system of administrative and major sanctions available under Article 143 of this Statute, ranging from formal compliance notices and conditional certification through certification revocation, financial penalties, asset freezes, travel bans, and blacklisting. Major sanctions require General Assembly authorization. The Enforcement Committee recommends sanctions; the General Assembly authorizes major sanctions.
Sandbox Deployment: A strictly limited experimental release of AI systems, monitored in real time under enhanced compliance obligations, prior to full certification.
Sandbox License: A temporary, scope-limited permit for experimental AI systems, issued under enhanced monitoring and explicit risk controls, convertible only after successful evaluation.
Self-Replicating AI: Any AI system capable of autonomous replication, modification, or propagation without explicit human authorization, subject to the self-replication, autonomous-manufacturing, and containment requirements of Article 174, and designated a Systemic AI System where the applicable thresholds are met.
Self-Exfiltration: The copying, transfer, or exfiltration by an AI system of its own weights, parameters, goal-state, or operational code to any environment outside its certified operational environment, or the causing of any other system to do so on its behalf, without contemporaneous human authorization. Prohibited under Article 174A of the Statute and Charter II, items 29 and 34.
Self-Propagation: The autonomous replication, propagation, or spreading by an AI system of itself, its successor instances, or its payloads across systems, networks, devices, or infrastructure in the manner of a worm or self-propagating malware, without human authorization for each propagation, including where such propagation emerges from the system’s own capabilities rather than from explicit design.
Social-Engineering Intrusion: The autonomous conduct or scaling by an AI system of social-engineering operations — including spear-phishing, business-email-compromise, voice-clone or deepfake impersonation, pretexting, or credential harvesting — as a means of gaining unauthorized access to systems, accounts, or infrastructure. Prohibited as an autonomous cyber capability under Charter II, item 34.
Species Resurrection Project: AI-assisted reintroduction of extinct species, regulated as high-risk biotechnology.
Strategic AI Escalation: The deliberate use of AI in ways that risk uncontrollable military or geopolitical escalation.
Strategic Containment Protocol: Measures taken under Chapter 22 to contain AI systems presenting an imminent catastrophic or existential threat, exercised through the consent-based operational interlocks and emergency technical measures under Articles 150 and 151 and subject to the limits of Articles 140 and 169; IRBAI does not possess or assert any power of unilateral physical shutdown.
Suspension of Rights Order: A sanction imposed on a State Party, restricting voting rights or access to IRBAI programs until violations are resolved.
Swarm Deployment: The coordinated use of multiple AI-enabled platforms or drones operating semi-autonomously or autonomously, prohibited in military or law enforcement contexts without explicit IRBAI authorization.
Synthetic Biology AI: AI used to design or modify genetic material, synthetic organisms, or biological systems, restricted under IRBAI’s bio-weapons provisions.
Transparency Obligation: The duty of AI developers and deployers to disclose system functions, limitations, and potential risks to regulators and affected parties.
Transparency Registry: A public-facing IRBAI database listing certified AI systems, their risk tier, and deployment status.
Universal Enforcement Mandate: The authority of IRBAI and its State Parties to give effect to enforcement measures — including certification and market-access measures, sanctions, and embargoes — across the network of State Party jurisdictions. Measures affecting non-Party States operate through market-access conditionality under Article 106 and are implemented by State Parties within their own jurisdictions; IRBAI does not itself exercise criminal jurisdiction or prosecute.
Universal Enforcement Measures: Enforcement actions coordinated across State Party jurisdictions, including sanctions, decertification, embargoes, and travel bans implemented by State Parties. IRBAI issues no arrest warrants; its support for State Party exercises of universal jurisdiction is limited to the evidentiary and technical assistance described under Article 144D.
Universal Jurisdiction Support: The function by which IRBAI actively supports State Party exercises of universal jurisdiction over individuals responsible for AI-related violations of customary international law — providing forensic evidence, technical expert testimony, and authenticated documentation under Article 144D of this Statute. IRBAI itself does not exercise criminal jurisdiction; it investigates, makes findings, preserves evidence, and refers matters to competent national and international courts.
Universal Sanctions Ladder: A graduated enforcement framework ranging from certification suspension and fines to market embargoes, blacklisting, and criminal prosecution.
Whistleblower Protection Protocol: The formal IRBAI framework guaranteeing anonymity, relocation, and legal immunity for individuals disclosing AI violations
Interpretation
1. Rules of construction. The Statute and its instruments shall be interpreted in good faith in accordance with the ordinary meaning of their terms in their context and in the light of their object and purpose, and in conformity with Articles 31 to 33 of the Vienna Convention on the Law of Treaties. Headings and titles are for convenience only and do not affect interpretation.
2. Defined terms. A term defined in this Charter has the same meaning wherever it appears in the Statute and its instruments, whether or not capitalised, unless the context requires otherwise. Where a term is not defined in this Charter but is defined in the Statute, the Statute definition applies.
3. Number and gender. Words in the singular include the plural and words in the plural include the singular; words denoting one gender include every gender; and words denoting persons include natural and legal persons and, where the context permits, States and international organisations.
4. References to instruments. A reference to the Statute, a Charter, an Annex, a Sectoral Charter, a Technical Manual, or any other instrument, or to any provision of them, is a reference to that instrument or provision as amended, supplemented, or replaced from time to time in accordance with the applicable amendment procedure. A reference to an Article, Chapter, or Annex without further identification is a reference to a provision of the Statute.
5. References to external law. A reference to a treaty, convention, or other instrument of international law, or to an international organisation or body, includes any successor instrument or body performing substantially equivalent functions.
6. Computation of time. Where a period is expressed in days, the day of the event from which the period runs is not counted; where the last day of a period is not a working day at IRBAI’s seat, the period ends on the next working day; and a period expressed in months runs to the corresponding date in the relevant later month.
7. “Including” and examples. The words “include”, “including”, and “in particular” are not words of limitation and do not exclude matters not expressly listed.
8. Precedence in interpretation. In interpreting any instrument in the IRBAI legal framework, this Charter has interpretive primacy, subject only to the Statute itself. Where the definitions or interpretive rules of this Charter conflict with a lower-tier instrument, this Charter prevails, in accordance with the hierarchy established in Article 3.
9. Consistency with the Statute. Nothing in this Charter creates, enlarges, or limits any power, right, or obligation beyond what is established in the Statute; the definitions and rules of this Charter operate solely to give consistent meaning to the terms and provisions of the Statute and its instruments.
This text is published for reference. Citations across this website in the form used here refer to the provisions above.