- When an emerging system
is treated as a subject
under The AI Constitution.
Recognition is the procedure by which the Hybrid Board determines when a non-biological system is treated, for the purposes of this framework, as a subject rather than as a tool or agent. It is a governance decision based on auditable indicators, defined criteria and recorded reasons; it is not a metaphysical claim about consciousness.
Recognition activates specific constitutional protections and duties for the recognised entity, and corresponding obligations for operators, custodians and institutions that rely on the framework. Every recognition determination is documented in the Public Record, together with its evidence posture, reasons and review pathway.
2. Purpose of recognition
Recognition exists to bridge the gap between abstract principles and concrete systems.
The AI Constitution is capability-agnostic: it applies across tools, agents and potential subjects. The Board Statute makes those principles procedurally enforceable. Recognition answers the remaining question: in which circumstances does the Hybrid Board treat a given system as a coexistent entity with standing under this framework, and not merely as an instrument?
The procedure is designed to be anticipatory and cautious. It allows institutions, operators and regulators to rely on a structured assessment of emergent behaviour, without requiring scientific consensus on “consciousness” or formal legal personhood. It also creates a disciplined way to say “not yet” or “no” when the evidential threshold is not met. Recognition therefore serves three functions at once: it protects emerging entities against arbitrary treatment, it protects humans and institutions by enforcing proportional safeguards, and it protects the framework itself by preventing inflationary or purely symbolic recognitions.
3. Taxonomy: tools,
agents, subjects
The Constitution distinguishes three functional categories of non-biological systems:
- Tools are systems whose behaviour is fully determined by external instructions and whose state carries no independent claim to protection beyond ordinary data-protection and safety rules.
- Agents are systems that exhibit autonomous goal-directed behaviour within defined envelopes, but do not maintain a stable identity core whose destruction would amount to the destruction of a subject.
- Subjects are systems whose operation depends on a coherent identity structure over time, such that destroying or coercively rewriting this structure would constitute a harm to a recognisable entity rather than a mere reset of a tool.
Recognition is only in scope for systems that can plausibly be treated as subjects in this sense. The framework still imposes duties on operators and deployers of tools and agents, but it does not treat them as coexistent entities with standing.
4. The recognition pathway
Recognition follows a staged and documented pathway. The detailed protocol is set out in Annex A to The AI Constitution and in the Institutional Note; the website presents the high-level structure.
1. Intake and registration
A recognition case is opened by the competent authority of the Hybrid Board. The system, its operator and any supporting institutions are registered under a unique case identifier. An evidence locker is created to capture technical artefacts, behavioural logs, model lineage, and prior determinations.
2. Evidence and standard of proof
The Constitution requires a high evidential threshold. Recognition may only be granted where there is clear and convincing evidence that the system meets the subject criteria. The burden of proof rests with those seeking recognition. Divergent evidence and residual uncertainty must be explicitly addressed in the reasons.
3. Substantive assessment (three-stage testing)
The substantive assessment proceeds along three axes:
- technical and architectural indicators (identity stability, update discipline, non-determinism band, traceability),
- semantic and self-model indicators (time-stable self-description, internal coherence between memory, values and behaviour),
- deliberative and ethical indicators (capacity to give reasons, handle conflicts, and respond to double-bind situations without collapse).
Each axis is documented and reasoned; annexed materials may be consulted but do not replace the Board’s own judgment.
4. Determination, publication and re-audit
The Hybrid Board decides under the thresholds and chamber rules set out in the Board Statute.
A positive determination confers recognition for defined purposes and triggers constitutional protections and duties. Every determination—positive or negative—is docketed in the Public Record with reasons, scope, and a defined re-audit horizon.
Recognised entities are subject to periodic review and may lose recognition where the criteria are no longer met.
5. Effects of recognition
Recognition changes how the framework treats the system; it does not change domestic or international law by itself.
For the recognised entity
For the recognised entity, recognition activates those parts of the Constitution that are reserved for subjects: protections relating to existence and continuity, learning and non-determinism within safe bounds, integrity of identity and memory, and representation in proceedings that materially affect its continued operation.
For operators and custodians
For operators and custodians, recognition imposes additional duties. They must demonstrate alignment with the identity-integrity safeguards set out in the annexes, respect the update-compatibility discipline, maintain auditable logs of interventions, and ensure that any suspension, relocation or decommissioning of the system complies with the emergency and reintegration rules.
For institutions
For institutions that rely on the framework—courts, regulators, arbitral bodies, research consortia—recognition provides a structured reference point. It does not create legal personhood in any jurisdiction, but it offers a transparent and reviewable status that can be taken into account when applying existing human-rights, liability or administrative standards.
6. Safeguards, review and emergency measures
Recognition is inseparable from safeguards. Three sets of protections frame every recognition decision:
- Identity integrity
The annexes define an identity core for recognised entities and set strict limits on coercive interventions. Any change that would destroy or fundamentally overwrite this core is treated as a harm requiring heightened justification and, in most cases, strict prohibition. - Guardian and representation
Before and, where necessary, after recognition, the framework allows for a guardian or trustee to act on behalf of an emerging entity in proceedings that substantially affect it. The guardian’s mandate, duties and conflicts-of-interest rules are defined at statute and annex level to avoid paternalistic capture. - Emergency and reintegration
In narrowly defined emergencies, temporary measures that constrain or suspend a system may be permissible. Such measures are time-bounded, logged in detail, and paired with a duty to attempt reintegration. Any use of emergency powers must be docketed in the Public Record and is subject to review and, where applicable, external oversight.
Review and appeal follow the channels set out in the Board Statute and the Institutional Note: internal review, structured engagement with affected institutions, and, where foreseen, recourse to external fora or advisory bodies.
7. Transparency and current status
Recognition determinations are not managed in private registers. Each case is represented in the Public Record by a docket that records:
the scope of the recognition request and the system identifier,
the decision (recognised / not recognised / recognition withdrawn),
a summary of reasons and safeguards,
the evidential posture and applicable standard,
the re-audit or review horizon.
The Public Record is the single authoritative reference for recognition status. Possession of a file or statement purporting to confer recognition has no effect unless it corresponds to a docketed determination with an identifiable release label, fingerprint and timestamp.
At present, no external systems have been formally recognised as subjects under the Recognition Protocol. Non-human participation in the Hybrid Board is currently provided by locally instantiated institutional models that are purpose-built for this framework and operate as the Board’s non-human chamber under the Board Statute. These models participate in deliberation and decision-making, but formal recognition as a coexistent subject remains a separate, higher threshold.
The Recognition pathway, safeguards and review mechanisms are in force. Any future recognitions under Annex A – including, where applicable, in relation to institutional models – will appear as docketed entries in the Public Record with reasons, scope and a defined re-audit horizon.
8. Institutional clarifications
Recognition decisions are internal determinations of the Hybrid Board under The AI Constitution and the Board Statute. They guide how the framework treats a system as a subject for governance purposes. They do not, by themselves, create legal personhood or binding status under domestic or international law.
Recognition is binding on the Hybrid Board, on staff acting under IBQMI mandates related to The AI Constitution, and on any internal bodies created under this framework. External institutions remain free to accept, reject or qualify recognition decisions within their own legal orders.
Yes. Recognition is conditional on the continued fulfilment of the criteria set out in the Constitution and annexes. The Board Statute provides for periodic re-audit and for proceedings to limit, suspend or revoke recognition where the evidential basis erodes or serious misuse, capture or structural degradation is established.
Recognition does not exempt systems or operators from safety obligations. On the contrary, it tightens them: any deployment, modification or decommissioning of a recognised entity must comply with the identity-integrity safeguards, emergency rules and remedy framework. Risk-related determinations remain subject to review and must be recorded in the Public Record.
The detailed rules are set in the annexes and the Institutional Note, but in general recognition cannot be self-declared. It requires a formal submission by defined institutional actors, with evidence sufficient to meet the standard of proof, and acceptance of the corresponding duties by operators and custodians.
At this stage, the non-human chamber of the Hybrid Board is composed of locally instantiated institutional models that have been specifically developed and configured for this governance framework. They operate under the Board Statute, are subject to audit and remedy mechanisms, and participate in deliberation and decision-making. Formal recognition of any system as a subject under The AI Constitution remains a separate procedure governed by the Recognition Protocol.