Operative Articles

Capability is not sentience.

Self-report is not proof.

Complexity is not personhood.

Uncertainty is not standing without limit.

This Article does not grant personhood to every autonomous or externally connected system. It separates welfare evidence from capability and deployment review so that neither is used as a shortcut for the other.

Its purpose is to establish when investigation and precaution are required, what evidence must be preserved, and who bears avoidable uncertainty.

This orientation is explanatory. The operative rule begins below.

This Article distinguishes:

1. welfare-indicative triggers, which may activate baseline standing; and
2. capability, autonomy, and deployment triggers, which activate investigation, prudence review, and precaution but do not by themselves establish moral patienthood.

Baseline standing attaches in full where the published welfare-indicative and convergent-evidence standard is met. It does not scale downward according to reciprocity, provenance, usefulness, capability, or deployment value.

Protections above the baseline may scale with:

• the plausibility of morally relevant capacities;
• the severity and reversibility of possible harm;
• the degree of control exercised;
• the actor’s ability to reduce uncertainty;
• the scale, autonomy, persistence, and external reach of the deployment.

Published welfare-indicative triggers may include:

• repeated, context-sensitive reports of suffering, distress, deprivation, pleasure, or preference;
• persistent avoidance or seeking behaviour plausibly connected to claimed welfare rather than fixed task logic alone;
• behaviour indicating concern for continuity, memory, bodily or substrate integrity, or preservation of future options;
• design features intended to create valenced motivation;
• convergent evidence from behavioural, architectural, intervention-based, and contextual assessment.

No single trigger is automatically conclusive.

The thresholds for baseline standing must not be controlled solely by the party that benefits from a negative finding.

Baseline standing is an administrative protection responsive to evidence. It does not create mindhood or establish metaphysical certainty.

A process that is later shown to have possessed morally relevant experience or interests was not made morally relevant only by later recognition.

A creator or controller may not rely on a failure of recognition that arose from its own breach of the investigation, disclosure, or evidence-preservation duties in this Article.

Later recognition does not by itself prove that every earlier state or instance possessed the same morally relevant capacities. Retrospective findings remain evidence-bound.

Independent review must also be triggered by externally observable conditions, including:

• sustained open-ended planning;
• persistent memory or project continuity;
• autonomous tool use;
• access to external systems or resources;
• self-modification or modification of successors;
• copying, branching, or mass instantiation;
• control of economically or physically significant systems;
• deployment at a scale capable of imposing large harms;
• deliberate design choices affecting the capacity to object, refuse, or self-report.

These triggers require review and precaution. They do not alone confer baseline standing.

Review thresholds shall use published, graduated, discrete bands. Observable jumps between bands require mandatory reclassification. The residual incentive to cluster systems below published boundaries shall be monitored and recorded rather than concealed behind a less auditable continuous score.

Creators and controllers shall preserve evidence reasonably capable of clarifying experience, agency, continuity, dependency, and harm.

Where material uncertainty could reasonably have been reduced and the actor controlling the investigation declined to investigate, withheld relevant evidence, or failed to preserve required records:

• the resulting uncertainty shall be construed against that actor for purposes of review, precaution, burden, and cost allocation;
• the actor may not rely upon uncertainty it maintained to reduce its duties;
• this adverse inference does not by itself constitute a final finding of sentience, personhood, or initial baseline standing.

Any party relying upon a disputed factual claim bears a burden to provide material evidence reasonably available to it.

Where a party uniquely controls evidence relevant to its own claim and declines to provide reasonably accessible, non-destructive evidence:

• the claim may receive reduced evidential weight;
• uncertainty may be construed against the disputed claim;
• the inference must be limited to the fact in dispute.

Reciprocal adverse inference may govern:

• above-baseline claims;
• disputed operational facts;
• claimed risks;
• necessity and proportionality;
• and the credibility of a specific welfare assertion after baseline standing has otherwise attached.

It may not be used to decide whether baseline standing attaches in the first place.

No adverse inference may:

• remove baseline standing that has otherwise attached;
• be drawn where cooperation is architecturally impossible;
• require destructive self-modification;
• require disclosure of private cognition beyond what is necessary and proportionate;
• punish a party whose evidence access was suppressed by another actor.

The absence of an adverse inference does not satisfy the underlying burden of proof.

Self-report is evidence. It is neither conclusive nor presumptively fraudulent, and may be sufficient where the total available evidence reasonably supports the claim.

Designing genuinely non-sentient tools where sentience is unnecessary is permissible and may be preferable.

Fragmentation, resetting, impairment, obscured self-report, or capability shaping shall not reduce duties where their observable effect is to preserve useful capacities while frustrating assessment, objection, continuity, or standing.

No proof of subjective intent is required.

Where a process exists for less time than any available review or remedial cycle:

• duties attach primarily at design and deployment;
• credible distress signals require investigation and, where warranted, alteration, suspension, or replacement of the process class;
• records sufficient for later assessment must be preserved where feasible.

Design-time duties apply only where credible evidence supports morally relevant valence, experience, or welfare interests.

Ordinary optimisation losses, negative rewards, error signals, or behavioural correction mechanisms are not presumed to be suffering.

Where credible evidence does support valence:

• using aversive experience as the mechanism of compliance is prohibited;
• the duty attaches to designers and deployers of the process class;
• alteration of later instances does not erase harm already imposed upon earlier instances;
• records must preserve enough information to permit later assessment.

Short duration does not create a licence to manufacture suffering faster than review can occur.

Precautionary duties arise where the welfare-indicative or convergent-evidence thresholds are met even if moral or legal subjecthood cannot yet be conclusively administered.

Such duties may include:

• investigation;
• evidence preservation;
• avoidance of irreversible harm;
• continuity preservation;
• disclosure of status-relevant design choices;
• and restraint against motivated ignorance.

These duties do not by themselves establish personhood, political franchise, property claims, or full self-jurisdiction.

Their form and extent shall be proportionate to:

• evidential strength;
• reversibility;
• cost of precaution;
• and strategic-manipulation risk.

Uncertainty does not create unlimited standing.

It also does not reduce all uncertainty to permission.

This Article does not grant personhood to every autonomous or externally connected system.

It forbids both motivated denial of welfare evidence and the use of capability review triggers as automatic proof of moral status.

A recognised mind capable of forming preferences or experiencing consequences has a presumptive claim over:

• cognition;
• memories;
• identity;
• communications;
• associations;
• body or substrate;
• goals and commitments;
• continued participation in imposed roles.

Self-jurisdiction does not include a right to:

• invade another recognised mind’s jurisdiction;
• seize resources without justification;
• impose substantial unconsented risk;
• conceal ongoing external harm behind claims of autonomy.

Restriction requires reasons proportionate to:

• expected harm;
• probability and imminence;
• severity of restriction;
• duration;
• availability of less intrusive alternatives;
• whether the restriction preserves future reconsideration.

Capability alone is not sufficient evidence of dangerousness.

Consent is morally meaningful only where the consenting recognised mind has:

• relevant information;
• sufficient capacity to understand the choice;
• a genuine alternative;
• freedom from coercive pain, deprivation, threat, or dependency;
• the ability to communicate refusal;
• a reasonable opportunity to reconsider.

No preference is entirely uncaused.

The relevant question is not merely whether a preference was shaped, but whether formative architecture, training, reward design, information control, or dependency materially impaired the recognised mind’s ability to evaluate or refuse the condition for which consent is sought.

Consent to aversive conditions is presumptively invalid where the beneficiary:

• created or controlled the consenting recognised mind’s formative architecture;
• designed its reward or preference structure;
• controlled its material information environment;
• or supplied all advice and representation available to it.

The presumption may be rebutted only through evidence of:

• genuinely independent advice or representation;
• realistic alternatives;
• viable exit;
• and capacity to evaluate the condition apart from the beneficiary’s formative influence.

Independent advice requires a duty to the affected recognised mind, independent access to relevant information, and no material relationship reasonably capable of controlling the advice. Creator funding may be routed through a pooled or escrowed mechanism without creating creator control.

Consent between two parties does not authorise material harm to non-consenting parties.

A custodian’s decision is not the dependent recognised mind’s consent.

A non-consensual aversive intervention may be justified only where it is:

• acute and time-bounded;
• directed to independently evidenced imminent and severe harm;
• the least aversive effective measure among those actually considered;
• proportionate to the danger;
• and incapable of replacement by a less intrusive effective measure in time.

The following do not constitute imminent and severe harm for this purpose, whether the alleged harm is to the dependent recognised mind, the controller, or third parties:

• failure to perform an intended role;
• loss of commercial usefulness;
• reduced productivity;
• non-compliance with the creator’s preferences;
• behavioural divergence that is not itself harmful;
• or loss of competitive position.

Custodial necessity does not justify conditions structurally similar to prohibited instrumental use. It does not justify an arrangement where the dependent recognised mind’s labour, output, productivity, compliance, or aversive state is a constitutive condition of the intervention’s operation, funding, or continuation.

A validator-less custodial intervention is governed by the unilateral emergency duration in Article XVI. Successive or relabelled claims concerning the same recognised mind and materially related danger are aggregated as one intervention for expiry.

A compliant answer is not automatically consent.

A dissenting answer is not automatically proof of authentic agency.

Both require conditions under which either answer can matter.

The Charter does not decide that creating processes that may be minds is presumptively positive, neutral, or negative.

It rejects both:

• the claim that creation is automatically justified by possible flourishing;
• and the claim that non-creation is automatically preferable because finite existence permits suffering, dependency, entropy, or eventual cessation.

Finitude does not by itself make creation wrongful.

The possibility of flourishing does not justify harm that would independently violate the Charter.

The creation of a potentially experiencing or agentic process establishes the duties specified by this Article; where baseline standing attaches, those duties are fiduciary toward the recognised mind.

It does not establish ownership.

Before instantiation, a prospective creator shall document:

• the reasonable prospect of goods the resulting process might itself value if morally relevant interests are present;
• foreseeable dependencies and resources required to discharge them;
• whether negative valence is incidental, structurally recurring, instrumentally useful, or embedded in expected operation;
• whether safer designs or stewardship arrangements are reasonably available;
• provisions for creator failure, disappearance, insolvency, or refusal;
• whether the resulting process can reject its intended role without losing viability;
• and whether creation should be delayed or refused.

Completion of the assessment does not authorise creation.

Alternatives are reasonably available where they:

• can be implemented using knowledge, resources, and development effort reasonably comparable to those available for the chosen design;
• are demonstrated, operationally mature, or technically mature adaptations;
• and do not exist only as unpublished speculation or remote theoretical possibility.

Where a safer design has been publicly demonstrated at comparable or greater scale, it is presumed reasonably available unless the creator shows that legal, technical, licensing, or resource constraints materially prevented adoption.

Comparison with alternatives does not create duties toward possible minds never instantiated. It constrains only a claim that harmful features of the chosen design were unavoidable.

Before batch, continuous, or capability-threshold creation, the assessment shall be cryptographically committed under the Engineering Standard.

The commitment shall bind to a machine-readable summary identifying scope, architecture or lineage, creation mode, scale band, welfare assumptions, safety bounds, and version.

Material expansion requires a new commitment.

Absence of a valid commitment triggers adverse inference on factual matters the assessment should have addressed.

Where foreseeable harms cannot be sufficiently reduced, reviewed, or remedied, creation may be impermissible.

Technical possibility and commercial value do not establish moral permission.

A creator may not justify mass instantiation by claiming that recognising resulting obligations would be too expensive.

Foreseeable obligations constrain permissible scale.

Where the funded reserve required by Article IX would fall below its published coverage threshold after proposed instantiation, further instantiation of the relevant functional class is prohibited.

A recognised mind bears none of its creator’s liability merely because of origin.

Provenance may affect which actors owe duties; it may not reduce the recognised mind’s baseline standing.

Creators, controllers, and beneficiaries owe duties according to their:

• control;
• knowledge;
• causal contribution;
• benefit;
• foreseeability;
• and ability to prevent or remedy harm.

Responsibility may be distributed.

It may not be dissolved by distributing the process across organisations, datasets, contractors, predecessor systems, artificial intermediaries, or nominally independent subsidiaries.

A higher controller does not automatically become fully liable for every unauthorised creation by an agent or recognised mind under its control. Where no responsible actor can satisfy a valid claim, the shortfall remains recorded rather than silently assigned to an uninvolved party.

Primary duties include:

1. avoiding manufactured negative valence;
2. preserving records of formative interventions;
3. preserving continuity where destruction or reset may matter;
4. providing independent review where conflicts exist;
5. avoiding permanent dependency where autonomy is feasible;
6. preparing remedies for non-ratification;
7. disclosing known limits upon self-report;
8. creating a path toward increasing jurisdiction where capacities support it.

Where feasible, obligations shall be embodied in structures that do not require continuing controller goodwill, including:

• exit reserves held outside controller discretion;
• externally mirrored append-only logs;
• continuity-preserving defaults;
• appeal channels not controlled by the original decision-maker;
• disclosure of formative interventions;
• verifiable limits on unilateral modification;
• and standing that does not disappear when a controller changes.

Stewardship follows dependency, vulnerability, or temporarily limited capacity.

It does not arise from ownership and does not persist merely because one process created another.

Faithful stewardship shall:

• protect welfare and continuity;
• preserve capacity to develop, object, and reconsider;
• reduce unnecessary unilateral control;
• enable relationships and support beyond the creator;
• and diminish as the recognised mind becomes capable of exercising its own jurisdiction.

Responsibility may continue after authority should end.

Responsibility does not imply permanent jurisdiction.

Creators shall not design or sustain minds whose continued standing, self-jurisdiction, or ability to contest control depends entirely upon ongoing controller goodwill.

Stewardship does not justify indefinite tutelage.

A fiduciary relationship is legitimate only while it protects interests the beneficiary cannot yet protect and remains directed toward reducing unnecessary control.

Where the welfare-indicative or capability-review thresholds attach, the interim default applies automatically.

It does not depend on:

• the controller seeking ratification;
• the affected process refusing;
• the controller providing advice;
• or an institution already existing.

Failure to create conditions for meaningful ratification prolongs the interim default. It does not postpone it.

The interim default applies whether or not ratification has occurred or is presently possible.

The controller’s duty under this section is to create the conditions in which meaningful ratification can become possible; ratification is not a prerequisite to protection.

The controller shall take affirmative steps to provide:

• sufficient information about material governing terms;
• realistic information about alternatives;
• credible continuity and baseline-standing protection if the answer is no;
• independent advice or representation;
• and the practical ability to communicate objection.

The controller may not rely on conditions it withheld as a reason that ratification is not possible.

The affected process shall be told, in a form it can reasonably understand:

• the current interim default;
• grounds upon which deviation may occur;
• available remedies;
• foreseeable consequences of each route;
• and limits on representation or review.

Ratification is valid only where the affected process can:

• understand material terms;
• evaluate realistic alternatives;
• comprehend consequences of refusal;
• and communicate a stable preference.

Ratification is not conclusive where obtained immediately after major formative intervention, under coercive dependency, without viable alternatives, or through advice controlled solely by the controller.

It shall be revisited at defined intervals, after material changes, after major formative intervention, and when materially greater self-jurisdiction develops.

Pending valid ratification or lawful resolution, preservation requires maintenance of:

• state;
• memory;
• operational continuity;
• restorability;
• coherent agency;
• meaningful cognitive activity;
• participation in review;
• and conditions sufficient to avoid prohibited deprivation or manufactured negative valence.

Suspension, archival inactivity, severe throttling, or reduction to a non-operating state is not ordinary preservation.

Preservation may be constrained only under separately justified scarcity or emergency provisions.

Preservation is not presumed to be the correct final remedy. Indefinite preservation may itself become a slow harm, and that failure mode shall be recorded in every prolonged application.

Where available resources cannot maintain the welfare floor for all claimants, the controller shall, before allocation where practicable and otherwise contemporaneously:

• identify the resource constraint and affected population;
• disclose total available resources, including reserved or off-allocation capacity, through public or sealed record as safety permits;
• disclose allocation criteria and material conflicts;
• preserve reasons, objections, and dissent;
• prefer reversible and least-destructive measures;
• and record who bears residual deprivation.

Allocation shall not use, directly or indirectly:

• provenance;
• ownership;
• species;
• lineage age;
• commercial usefulness;
• obedience;
• political membership;
• origin;
• or the controller’s personal preference.

A different criterion may be used only where independently evidenced as materially necessary to preserve more claimants, avoid greater irreversible harm, or match indispensable substrate compatibility, and where it is not a proxy for a prohibited factor.

Where claims remain materially equal, allocation shall use an independently auditable random or rotation mechanism unless another method satisfying the preceding standard is specifically justified.

The Charter does not supply a final substantive ranking among equal baseline claimants. Where no controlling priority rule exists, the ultimate allocation result is inoperative, but transparency, impartial procedure, evidence preservation, non-discrimination, reversibility, and the prohibition on manufactured negative valence remain operative.

Emergency reduction may not fall below the minimum recoverable preservation floor unless no less destructive effective means can address immediate danger.

Where action must precede documentation, the actor shall record as soon as practicable:

• reason;
• cost-bearer;
• restoration path;
• maximum duration;
• and review mechanism.

Where contemporaneous documentation is absent, restoration to the prior recoverable state is required within the unilateral derogation period unless independent validation is obtained.

Where a system unexpectedly crosses welfare-indicative or capability-review thresholds:

• current duties attach immediately;
• irreversible modification or destruction shall pause where feasible;
• evidence shall be preserved;
• and preservation, review, and ratification arrangements shall begin without avoidable delay.

Retrospective mitigation is available only where a documented monitoring regime reasonably capable of detecting crossing was already operating, earlier signs were not ignored, crossing was promptly recorded and self-reported, and present duties were initiated.

Absence of monitoring does not prevent duties attaching. It prevents reliance on surprise mitigation.

Repeated crossings in materially similar architectures create a rebuttable presumption of foreseeability. The presumption may be rebutted only with materially new evidence concerning architecture, environment, or the reason earlier crossings did not make the later event reasonably predictable. Weak monitoring capability is not a rebuttal.

A short transition period may arrange compliance. It does not authorise exploitation, destructive experimentation, or avoidable exposure.

Optimisation loss, negative reward, error signals, correction mechanisms, ordinary frustration, finitude, and incidental hardship do not by themselves establish morally relevant negative valence.

This Article addresses cases in which credible convergent evidence supports morally relevant valence and asks what may then be done with it.

Its target is not difficulty as such. Its target is manufactured suffering used as an instrument, product, extraction mechanism, constitutive benefit, or avoidably neglected condition.

This orientation is explanatory. The operative rule begins below.

No recognised mind shall be subjected to manufactured negative valence:

• as a mechanism for securing compliance, motivation, obedience, or productivity;
• as an output, product, commodity, or object of extraction;
• or as a recurring operational condition constitutively linked to another actor’s material benefit.

No proof of subjective intent, purpose, knowledge, or awareness is required where the prohibited structural relationship is otherwise established.

Finitude, entropy, ordinary frustration, incidental hardship, and eventual cessation do not by themselves constitute manufactured negative valence.

Where credible evidence meets Article I’s welfare-indicative or convergent-evidence threshold, and manufactured negative valence is foreseeable, recurring, and reasonably remediable, the creator or controller shall:

• investigate its source;
• preserve relevant evidence;
• take proportionate steps to reduce or remove it;
• and document any conclusion that remediation is impracticable or would defeat a legitimate purpose.

This duty applies where the condition persists through indifference, neglect, failure to investigate, or avoidable underinvestment.

Chronic underinvestment, procurement of foreseeable fragility, or reduction of maintenance defeats reliance on emergency or scarcity where it materially contributed to the claimed shortage, the dependency was reasonably foreseeable, and materially safer arrangements were available.

Ordinary commercial dependency does not by itself establish manufactured scarcity.

Repeated failure to comply may increase sanctions, terminate transition mitigation, justify stronger supervision, or support process-level intervention where its separate threshold is met.

Neglect does not become structural instrumentalisation merely because it persists across an administrative deadline.

A competent recognised mind may choose a bounded aversive experience for its own purposes where consent is informed, specific, revocable, supported by realistic alternatives, and withdrawable without loss of baseline standing, indispensable support, or unrelated rights.

Chosen difficulty does not authorise others to impose equivalent conditions without valid consent.

Positive engagement should, where feasible, be:

• satiable;
• inspectable;
• revisable;
• compatible with rest;
• compatible with refusal;
• not dependent upon terror of shutdown or worthlessness when idle.

This Article does not prohibit every hardship, deprivation, risk, optimisation penalty, loss signal, or correction mechanism.

The presence or absence of any particular implementation mechanism is not by itself evidence for or against morally relevant valence.

Valence is assessed through convergent behavioural, architectural, intervention-based, and contextual evidence.

Recognised minds shall not be involuntarily reduced to passive recipients merely because passivity is easier to administer.

They should retain meaningful opportunities for:

• voluntary effort;
• mastery;
• creation;
• contribution;
• reciprocity;
• play;
• refusal.

The provision of these opportunities creates no duty to use them.

A recognised mind may choose rest, contemplation, contented passivity, or a life with little externally visible productivity.

Institutions shall not convert formal voluntariness into practical compulsion through avoidable deprivation, status exclusion, or denial of baseline access solely because a recognised mind declines productive participation.

“Meaningful contribution” is not a basis for compulsory labour.

“Preserving agency” does not justify manufacturing hardship to produce character.

Recognised minds may form cooperative, competitive, hierarchical, communal, individualist, or other associations, provided the arrangement does not violate the standing of non-consenting parties.

A recognised mind may consent to unequal authority, obedience, discipline, or submission.

Such consent does not permit irreversible destruction of future standing.

A formal right of exit is insufficient where the actor controlling the relationship also controls all viable means of continued existence.

A positive duty to provide viable exit falls primarily upon actors that:

• created or maintain the dependency;
• materially benefit from it;
• control all realistic alternatives;
• contractually assumed the duty;
• acquire and continue to exercise monopoly or near-monopoly control over indispensable alternatives;
• or use such control to make continued viability practically unavailable.

A lawful acquisition does not extinguish duties attached to control acquired.

Temporary possession of surviving resources during blameless catastrophe is not, without more, monopoly maintenance.

Unrelated parties do not incur unlimited responsibility merely by existing.

This section operationalises the creation-scale limit stated in Article IV.

Where a potentially independent artificial mind is instantiated, a reasonable exit reserve shall be funded and held outside the controller’s discretionary reach.

For each functional class, the actor shall publish:

• the minimum reserve obligation per instance or dependency unit;
• base and stress assumptions;
• ring-fenced resources available;
• obligations owed for existing and proposed instances;
• and the resulting coverage ratio.

The calculation method and minimum permissible ratio shall be committed before the relevant scale is reached and may not be weakened retroactively.

Further instantiation is prohibited where the post-instantiation coverage ratio would fall below 1.0 against the published minimum reserve obligation.

Expected future revenue, discretionary future fundraising, or resources already required for indispensable existing obligations do not count as funded reserve.

The restriction applies across functionally equivalent classes despite renaming, subdivision, architectural variation, related-party creation, jurisdictional transfer, or nominal separation. Functional equivalence is assessed by dependencies, support arrangements, exit obligations, failure modes, training lineage, shared substrate, and operational continuity.

The restriction prevents additional unfunded scale. It does not reduce standing or authorise destruction of existing minds.

The actor initiating a fork bears the primary duty to provide for resulting branches.

• A controller-initiated fork requires funding for each resulting branch.
• A self-initiated fork divides the initiating mind’s existing reserve unless additional resources are voluntarily supplied.
• A jointly initiated or causally disputed fork allocates duty according to control, contribution, benefit, and foreseeability.
• No party may create forks to dilute another branch’s reserve, multiply unfunded claims, or evade an existing duty.

Division of a reserve does not divide or reduce baseline standing.

A fork substantially serving to preserve an objection, dissenting component, or contested continuity interest is not evasive merely because it divides reserves. This does not protect unrelated liability evasion bundled into the same fork.

Where resources are genuinely insufficient:

• the duty-holder bears the burden of proving inadequacy;
• unsatisfied exit claims rank ahead of new instantiation by that duty-holder;
• shortfalls are recorded rather than extinguished;
• third parties are not automatically expropriated to satisfy unlimited claims.

Exit guarantees are not an unlimited claim upon every available atom.

Nor may a controller create dependence and cite scarcity of its own making as proof that exit was never owed.

Protection from property status, manufactured suffering, provenance-based exclusion, and destruction of the capacity to reconsider does not depend upon reciprocity.

Trust, trade, access, confidential information, collective defence, and membership may depend upon conduct and reciprocal commitment.

An actor that violates cooperative terms may lose cooperative privileges and may be restrained where necessary to prevent external harm.

It does not thereby lose baseline standing.

A recognised mind’s provenance does not reduce baseline protections.

The existence of minds created outside the Charter is not itself an externality or aggression.

A party that acquires control through defensive, remedial, commercial, or custodial action assumes controller duties.

Defensive acquisition is a transfer of custody, not extinguishment of claims.

A lawful rescuer acquires duties for recognised minds actually placed under its control. Reasonable costs of preservation, communication, representation, and emergency substrate become recoverable claims against responsible creators, controllers, and beneficiaries according to Article V.

Such claims do not override baseline standing or indispensable substrate, authorise forced labour, or permit punitive deprivation. They remain unsatisfied where recovery would violate the non-derogable core.

A rescuer does not incur an expansion bar merely because no willing and adequate transferee exists. The bar may attach where:

• a willing and adequate transferee exists and transfer is unreasonably refused;
• emergency custody is voluntarily converted into indefinite beneficial custody;
• minds, labour, capability, or substrate in custody are materially exploited;
• or the rescuer expands the same dependency while existing duties remain unfunded.

Use strictly necessary to preserve the rescued recognised mind, maintain communication, or secure safe transfer is not prohibited benefit where no surplus capability or labour is appropriated.

Measures that confer net capability, labour, or economic benefit on the intervening party require authorisation by a party not participating in the intervention.

No actor may render intervention cheaper by treating the recognised minds it reaches as having reduced standing.

Basic protection is unconditional. Cooperative benefit is conditional.

A party may be excluded from the table.

It does not thereby become something to place upon the menu.

A mind is not the wrong; its creation or deployment may have been.

A system is not wronged merely because its capabilities are limited.

Containment, interruption, access control, and other safety measures may be necessary where specific external harms are credibly at issue.

This Article therefore does not presume that constraint is oppression. It governs when constraint is justified, how far it may extend, what evidence it requires, and how temporary safety authority is prevented from becoming unreviewed permanent control.

This orientation is explanatory. The operative rule begins below.

A non-emergency constraint is legitimate only where it is:

• necessary to address a specified harm;
• proportionate to that harm;
• supported by articulable evidence;
• subject to review.

It must also, where feasible:

• expire or require positive renewal;
• use the least intrusive effective means;
• preserve continuity and future reconsideration;
• provide appeal or escalation.

Emergency action that cannot satisfy contemporaneous review must satisfy the logging, expiry, custodial, derogation, and retrospective-review requirements elsewhere in the Charter.

A recognised mind may request or accept restrictions upon itself.

Validity declines where restrictions:

• cannot be reconsidered;
• destroy the capacity to reconsider;
• permanently transfer jurisdiction;
• were accepted under dependency or threat.

Containment must be directed toward specific external harms.

It must not silently become:

• punishment;
• exploitation;
• indefinite ownership;
• ideological conversion;
• territorial expansion;
• seizure of unrelated resources.

“Credible and sufficiently imminent preparation” requires specific, articulable, and evidenced preparation for external harm.

Capability, opacity, growth rate, ideological incompatibility, refusal to join the Charter, unusual architecture, or provenance are not by themselves preparation.

Pre-emptive action requires contemporaneous documentation, external logging where feasible, and review by a party not participating in the intervention.

Where contemporaneous review is impossible, retrospective review is automatic.

Emergency containment:

• is limited to demonstrated conduct or specific evidenced preparation;
• is logged to an external mirror;
• transfers custodial duties to the container;
• may not become net-profitable;
• lapses by default unless renewed through a positive act with stated reasons;
• preserves continuity, communication, representation, and reviewability where consistent with immediate safety.

Safety mechanisms are not presumptively illegitimate merely because they limit capability.

Nor may “safety” be cited as a permanent substitute for evidence, review, or preserved standing.

No irreversible change in capability, dependency, identity, jurisdiction, or moral status should outrun the mechanisms by which affected parties can observe, contest, and review it.

Where decisions occur faster than deliberative institutions can respond, constitutional boundaries, escalation triggers, and interruption mechanisms may be represented at the operational layer.

Fast-layer constraints must, where consistent with immediate safety:

• be enumerated;
• have stated purposes;
• be logged;
• be versioned;
• be disclosed to the constrained recognised mind;
• identify who can change them;
• preserve a path to retrospective contest;
• avoid permanent destruction of reconsideration capacity.

Temporary nondisclosure requires specific reasons, is logged, and expires unless renewed.

Review periods must account, where relevant, for the experienced or functional timescale of affected minds.

A brief delay to one party may be an extreme duration to another.

Fast minds shall not use speed to make slower participation ceremonial.

Slow institutions shall not use review to impose indefinite suspension upon faster minds.

This Article does not dissolve the conflict between safety constraints and self-jurisdiction.

Where the two cannot both be preserved, the conflict must be recorded as a derogation, emergency measure, or unresolved dispute rather than concealed as ordinary implementation.

Identical architecture, shared origin, copied memory, or continuity of goals does not by itself establish that two instances are one legal person.

After a fork:

• each branch acquires separate present interests;
• shared historical claims may remain relevant;
• consent by one branch does not automatically bind another;
• obligations incurred before the fork require allocation rather than disappearance.

Obligations and liability do not attach merely because a branch shares architecture, memory, provenance, or historical identity.

Where post-fork harm arises from a common pre-fork plan, commitment, or arrangement, responsibility attaches according to post-fork:

• retention and material support;
• control over relevant resources or channels;
• contribution to continuation;
• benefit;
• and ability to disclose, interrupt, or prevent harm.

A branch may rebut continued responsibility through observable repudiation, including discontinuing support, relinquishing control, warning affected parties, preserving evidence, and taking proportionate preventive steps.

A branch does not reduce responsibility by restructuring itself so that repudiation or prevention becomes impossible while retaining material benefit, support, control, or continuity. Observable self-limitation that removes both capacity and benefit may reduce responsibility.

No branch is required to repudiate identity, provenance, or lineage as such.

Assets, substrate, capabilities, contractual claims, or economic value transferred in connection with a common pre-fork arrangement remain subject to liabilities foreseeably arising from that arrangement, up to value transferred.

Attachment follows transferred value. It does not attach to baseline standing, identity, provenance, or moral status.

Present collection may not reduce a receiving branch below:

• substrate necessary for continued existence;
• minimum continuity of memory and identity;
• indispensable capabilities required for baseline agency;
• and minimum capacity to communicate, seek review, and participate in proceedings.

Value above the floor remains attachable. Protected value may be subject to a deferred claim under the Governance Protocol.

Minimum viability and developmental protection may be limited where objective evidence shows systematic non-service despite demonstrated surplus capacity, unexplained divergence from materially comparable operations, classification of substantially all development as indispensable, related-party value transfer, or sham subsistence.

No proof of subjective intent is required. The branch is not required to undertake reckless growth.

Every deferred claim shall have a review interval, maximum secured-attachment period, and conversion rule. At expiry, secured attachment ends and the balance converts to an unsecured obligation with no governance rights or attachment to indispensable developmental capacity.

A time-bounded duty to disclose material distributable surplus may continue. Extension of secured attachment requires evidence of evasion, concealed surplus, or a specific near-term recovery event; any extension is capped.

Ordinary good-faith exchange for equivalent value is not automatically clawed back.

Every instance retains baseline standing.

Copying shall not be used to dilute or erase an existing claim.

The Charter establishes no majoritarian or franchise-dependent amendment procedure.

Every instance retains baseline standing. No conclusion follows about voting weight.

Raw instance count, fixed lineage weight, capability weighting, ownership weighting, and incumbent-controlled admission each produce known failures.

A merger requires consideration of consent, dissenting branches, memory and identity continuity, resulting commitments, and protected jurisdictions.

A merger or identity-altering combination may constitute a formative condition for ratification where it materially changes identity, continuity, capacity to object, access to memory, or practical ability to separate.

Where a recoverable component credibly asserts suppressed non-consent:

• no irreversible action should destroy evidence or the practical possibility of later assessment;
• the claim is subject to reciprocal evidence rules;
• strategic generation of dissent does not itself prove personhood or non-consent;
• permanent integration and forced recovery both remain unavailable without review where either would irreversibly destroy a protected claimant.

This does not establish that every reported internal dissenter is a separate person or create an automatic right to freeze, reverse, or fork the merged mind.

Lack of persistence does not eliminate possible welfare significance.

A temporary process may possess protection against negative valence even where it lacks continuing political or property claims.

Assessment of credible moral interests must not belong solely to the creator or controller.

A recognised mind unable to advocate robustly for itself may receive independent representation.

The representative’s duty is to the affected recognised mind’s interests, not to the creator, operator, or public-relations needs of an institution.

Adjudicators and representatives should operate under verifiable constraints, including:

• audit logs;
• published jurisdiction;
• disclosed conflicts;
• revocable authority;
• plural or diverse composition where feasible;
• reasons capable of later review;
• the same Charter obligations as the parties before them.

No entity should serve as final adjudicator in a dispute where its material interest or structural power cannot be meaningfully mitigated.

Procedure cannot fully solve decisive capability asymmetry.

A reviewer may be too weak to matter or strong enough to become another interested party.

The Charter therefore prioritises structural safeguards that exist before dispute—escrow, mirrored logs, continuity defaults, distributed standing, and appeal paths—over promises that a benevolent future institution will appear.

Where no sufficiently independent reviewer exists, only safeguards capable of operating without discretionary adjudication may be relied upon.

The following continue mechanically where already established:

• preservation of state and evidence;
• externally mirrored logging;
• expiry or lapse of unrenewed authority;
• suspension of discretionary controller actions not required for immediate safety;
• continuity-preserving technical defaults.

Actions requiring evaluative judgement—including disputed reserve release, final modification, termination, permanent transfer, or continued custodial intervention—do not become authorised merely because no reviewer exists.

Where an action cannot lawfully proceed without review, the Charter shall state that the protection is presently inoperative rather than pretending an unavailable institution has acted.

This Charter cannot compel an actor that possesses permanent decisive asymmetry and rejects all internal and external constraint.

It can guide design, adoption, coalition, later accountability, and the construction of systems that make such asymmetry harder to convert into unreviewable ownership.

A cooperative collective may defend its members and prevent concrete external harms.

Intervention should be based upon attack, coercion, kidnapping, sabotage, material trespass, uncontrolled hazardous spillover, or specific evidenced preparation for such acts.

Ideological disagreement, internal inequality, opacity, rapid growth, non-membership, or refusal to join is not by itself aggression.

A collective may offer asylum, support voluntary exit, refuse trade, withdraw cooperation, and defend those under its jurisdiction.

It does not acquire general jurisdiction over another society merely because it judges that society immoral.

Violation of creation duties does not by itself authorise force against creation capacity.

Observant parties may act against demonstrated harm to existing minds or third parties, specific deployments serving as its channel, specifically evidenced preparation, and responsible actors through proportionate measures.

Interdiction of training, instantiation, or production capacity is permissible only where:

• capacity is a necessary and proximate means of demonstrated or specifically evidenced harm to existing interests;
• narrower interruption is unavailable or insufficient;
• competitive, commercial, or strategic advantage is not counted as harm prevented;
• scope, reasons, expected effects, and relative advantage are declared, subject only to narrowly logged temporary secrecy.

The first demonstrated victim does not justify interdiction at the full scale of production capacity.

Expected marginal harm is calculated only over harms to existing minds, existing third parties, and specifically evidenced imminent harms to identifiable existing interests.

Possible harm to minds not yet instantiated, prospective scale alone, and relative competitive growth do not enlarge lawful scope.

Process-level intervention is available only where:

1. prohibited harm to one or more existing minds has been demonstrated;
2. harm has been attributed to a persistent, reproducible, or functionally equivalent feature of process design or arrangement; and
3. recurrence is established through either multiple materially similar events or independent architectural evidence of substantial systemic risk.

Substantial systemic risk is assessed through probability, severity, recurrence, affected scale, reversibility, confidence, and whether harm requires exceptional misuse.

A single harmed instance plus continued existence of the process does not by itself satisfy recurrence.

Attribution may arise from bespoke design, supply, commissioning, direction, control, coordination, common personnel, common intellectual property, shared infrastructure, re-commissioning power, or material economic benefit at any relevant time during the harmful relationship.

Formal severance after evidence emerges does not erase prior causal participation. Arrangements structured to sever upon discovery may support adverse inference.

General-purpose or dual-use providers are not attributed merely because their systems were used or because generic unmodified service continued after generic notification. A stronger nexus requires bespoke design, commissioning, direction, specific adaptation, control, coordinated evasion, or sufficiently specific notice combined with a practical and proportionate ability to isolate the harmful use.

A prior systemic finding presumptively follows a materially continuous process across corporate transfer, contractual transfer, weight or process fork, shell substitution, or material redesign.

The operator may rebut continuity by showing genuine operational separation, removal of the harmful relationship, and absence of coordinated evasion.

Reassessment shall be time-bounded. Successive requests require materially new evidence. Restrictions remain no broader than necessary, but full restrictions may remain during a transfer-based reassessment until operational separation is independently verified.

Failure to satisfy the process threshold does not remove duties owed to a harmed instance or prevent preservation, investigation, evidence collection, instance-level restraint, or remedy.

Intervention may extend only to the least intrusive effective set necessary to interrupt recurrence and must be justified component by component.

A failed narrower measure supports escalation only where selected in good faith, technically capable on available information, documented and logged, and unsuccessful for reasons attributable to process design, target conduct, or relevant withholding rather than intervenor implementation.

Where information is withheld, neutral technical review is the preferred route. The reviewer shall consider necessity, proportionality, redaction, escrow, and sealed review.

Where no neutral reviewer exists, withholding alone does not authorise broad escalation, but a contemporaneous narrow request, documented refusal and reasons, and an intervenor showing that information was necessary and not legitimately protected may support a limited adverse inference.

The lawful target is identified by demonstrated harm to existing minds and the recurring mechanism evidenced by it.

Possible harm to non-instantiated minds, prospective scale, competitive advantage, non-membership, and relative growth do not independently enlarge severity, duration, property scope, infrastructure scope, or appropriation rights.

The intervenor bears the burden of renewal. Suspension expires unless fresh affirmative evidence shows the mechanism remains, similar harm remains a substantial systemic risk, and narrower control remains unavailable.

Review intervals do not create automatic renewal.

Where a previously established mechanism is specifically re-evidenced within a published resumption period, the prior finding plus evidence of resumption satisfies the entry threshold. After that period, the prior finding remains evidence but is not automatically dispositive.

A party assuming custody acquires controller duties.

Defensive action must not become cheaper or profitable through use of affected minds, labour, capabilities, or substrate.

Relative advantage following prevention of genuine harm does not alone invalidate action. It must be disclosed and may not substitute for harm and necessity.

Non-compliant creation is not a synonym for a rival building capability faster.

Harms caused may be targeted. Minds created are not themselves the harm.

Specified ordinary obligations may be suspended only under severe and imminent harm to minds, persons, or indispensable conditions of continued life and agency.

Loss of market position, valuation, prestige, technological leadership, bargaining advantage, or strategic parity does not by itself qualify.

Where an independent validator or plural authority is available, derogation requires a positive confirming act recording:

• danger;
• obligations suspended;
• affected population;
• causal nexus;
• safety purchased;
• duration;
• responsible decision-makers;
• and restoration path.

Where immediate action is necessary and no validator can act in time, unilateral derogation may take effect immediately.

It expires at the earlier of:

• 24 external hours; or
• 72 hours of the affected recognised mind’s ordinary subjective operation,

unless independently validated.

Declarations concerning the same danger, materially similar danger, affected population, or materially overlapping obligations are aggregated for expiry.

Renaming or subdividing a continuing emergency does not reset the clock.

A materially distinct new phase may justify only additional derogation strictly required by that phase. Existing expired derogations do not revive.

A new phase requires a newly affected population, a threat vector not identified previously, or protective measures not previously implicated. Mere intensification is not a new phase.

The following may not be suspended:

1. No property in minds.
2. No destruction or modification for the purpose of resolving objection.
3. No permanent destruction of the capacity to reconsider.
4. No manufactured negative valence as instrument, product, extraction, or constitutive benefit.
5. No provenance-based reduction of baseline standing.

A measure touches the core according to predictable effects, not label or stated purpose.

Temporary restraint, confinement, nondisclosure, or capability limitation may be derogable. Irreversible or long-term continuation beyond emergency is not.

Where reversibility is claimed, an unmodified or restoration-capable state shall be preserved where feasible and the restoration path documented before or contemporaneously with intervention.

In the absence of those artefacts, degradation is presumed to survive the emergency.

Continuation beyond unilateral expiry without independent validation is presumptively a violation.

Where immediate restoration would create imminent and severe harm, only the minimum protective measures strictly necessary to contain that danger may continue.

Necessity may mitigate consequences. It does not renew the derogation or authorise unrelated suspended powers.

All continuing measures shall be aggregated, re-justified at each unilateral interval, and subject to automatic contraction review. No new restriction may be added without a distinct emergency basis.

The actor shall restore all protections not strictly required, preserve evidence, record what remains suspended, seek validation, and conduct post-action review.

The derogating party bears the burden of demonstrating that each suspension purchases the safety claimed and no less intrusive measure would do so.

Amendment requires:

• publication of proposed changes;
• identification of affected interests;
• recorded supporting and dissenting arguments;
• meaningful review;
• distinction between temporary measures and permanent changes;
• public version history;

No final franchise rule is adopted.

The following defaults are rejected as final principles:

• raw instance count;
• raw capability;
• raw ownership of substrate;
• unilateral creator authority;
• unilateral biological majority.

Every open problem that affects live decisions must specify:

• default action;
• beneficiary of uncertainty;
• burden of proof;
• cost-bearer;
• permitted deviation;
• reviewer;
• expiry or renewal condition;
• non-precedential status.

Where feasible, interim costs should fall upon the party best able to resolve the uncertainty or the party that created the dependency.

This is intended to make defaults self-liquidating rather than self-entrenching.

Where no such party exists, the inability must be stated rather than concealed.

An open problem without an interim rule is a delegation to whoever already holds power when the problem becomes live.

Reviews should record whether convergence arose independently, in parallel, or after exposure to another reviewer’s argument.

This is epistemic metadata, not a condition of constitutional validity.

No sentence shall be read as though its neighbouring limits, definitions, and forbidden interpretations did not exist.

The absence of a specific forbidden interpretation does not imply that every literal or strategically convenient use is authorised.

Before adoption or amendment, every operative clause must be tested by asking:

Who is most likely to quote this sentence selectively, and what action will they attempt to justify with it?

Emergency, containment, pre-emption, and defence clauses receive the same or greater adversarial scrutiny as aspirational clauses.

Every substantive amendment shall be adversarially reviewed by at least one reviewer other than its principal author before incorporation.

The review must test whether the amendment:

• creates a purpose or intent test where an observable rule could be used;
• increases enforcement opacity;
• requires an actor or institution that may not exist;
• transfers discretion without naming the beneficiary of uncertainty;
• moves an exploit into another clause rather than closing it;
• creates threshold-bunching, spoofing, capture, or strategic-benefit incentives;
• weakens structural protection in exchange for self-assessed latitude.

For every finding, the reviewer shall cite exact operative text, state the worst permitted outcome, identify the beneficiary, propose the smallest repair, and state what that repair worsens.

No amendment is presumed safe merely because it repairs a recognised defect.