Who granted the power?
Authority Cloud traces power to an accountable source and preserves scope, purpose, limits, expiry, attenuation, and revocation. A valid grant is necessary, but it does not by itself prove sufficient consequence capacity.
Consequence Capital · Cross-platform primitive
No machine may create
an unreserved consequence.
A machine can be correctly identified, explicitly authorized, and technically capable of acting while the organization is still unprepared for what that action could create. StableMind calls the capacity required to absorb, contain, recover from, compensate for, or settle that exposure Consequence Capital.
PROPOSED ACTIONpayment.transfer
value: USD 250,000
destination: approved supplier
irreversibility: elevated
FINANCIALRESERVED
CONTRACTUALBOUNDED
OPERATIONALRECOVERY READY
LEGALSEPARATE DUTY
IRREVERSIBILITYNO NETTING
RELEASE GATEPROOF REQUIRED
RESERVE POSTURELOCKED
Capacity is committed to this exact action envelope and remains unavailable for reuse until governed release conditions are satisfied.
Synthetic architecture illustration. It is not a customer reserve, financial account, insurance policy, production payment, or external proof.
Category definition
What is Consequence Capital?
Consequence Capital is explicitly bounded capacity reserved before a consequential machine action so the organization does not discover its exposure only after execution. Depending on the action, that capacity may involve money, recovery capability, human review, legal or regulatory duties, privacy containment, security response, operational redundancy, contractual remedies, resource ceilings, environmental controls, or treatment of irreversibility. StableMind does not collapse those dimensions into one universal risk score. They remain non-fungible because a financial reserve cannot erase human harm, an insurance policy cannot substitute for authority, and a favorable model cannot make an irreversible action reversible.
The missing layer
01 / PERMISSION IS NOT PREPAREDNESS
Authorization answers “may the machine act?” Consequence Capital asks “are we prepared for what this action can create?”
Is this exact request inside the boundary?
ActionGate evaluates authority, policy, separation of duties, revocation, and required consequence conditions before an Action Permit may exist. Missing required reserve can narrow, escalate, or deny the request.
What capacity is committed before action?
Required consequence dimensions are bounded in their own native terms, assigned accountable ownership, locked to the action envelope, and conserved so the same capacity cannot be promised twice.
When can committed capacity move again?
Evidence can support the governed release decision after independent consequence verification. An executor receipt, optimistic forecast, dashboard state, or commercial payment cannot silently release reserve.
Reserve model
02 / SEVEN STATES
Consequence capacity should be conserved through the action lifecycle, not treated as an after-the-fact score.
AVAILABLE
Capacity exists and is not currently committed to a governed action. Availability is not authority and does not create permission to consume it.
DELEGATED
A bounded slice is made available to a specific authority envelope without being duplicated across descendants or tenants.
LOCKED
Capacity is committed to an exact action, target, limit, validity window, and consequence contract before execution begins.
CONSUMED
Execution has created or realized some portion of the governed exposure. Consumption remains attributable to the originating action.
DISPUTED
Evidence, settlement state, affected-party claims, or downstream conditions conflict. Capacity remains encumbered while the conflict is unresolved.
FROZEN
Unknown, stale, contradictory, unsafe, revoked, or otherwise unresolved conditions prevent release or reuse.
RELEASABLE
Governed release conditions have been satisfied. The interface may display this state, but it cannot itself release reserve or manufacture Proof of Consequence.
Non-fungible dimensions
03 / TEN CONSEQUENCE DOMAINS
There is no honest universal “AI risk balance.” Different consequences require different capacity.
StableMind preserves ten ontology-bound consequence dimensions from the underlying Consequence Capital model. Each dimension can carry its own ceiling, reserved amount or bounded condition, observed state, uncertainty, accountable owner, affected scope, settlement deadline, and release gate. Cross-dimension netting is prohibited where it would hide the nature of the exposure.
Financial
Monetary exposure, payment value, credit use, liquidity, recovery funds, penalties, settlement obligations, or other explicitly financial capacity.
Human
Consequences affecting people, safety, welfare, employment, access, dignity, or other human interests. Human consequence is not converted into a price merely to make arithmetic convenient.
Legal & regulatory
Jurisdictional duties, approvals, reporting obligations, supervisory holds, regulated limits, or legally required response capacity.
Operational
Redundancy, rollback, recovery windows, service continuity, failover, staffing, or other capacity required to contain operational disruption.
Data & privacy
Disclosure boundaries, deletion or recovery duties, retention controls, affected-person obligations, data residency, and containment capability.
Security
Credential revocation, isolation, incident response, blast-radius limits, quarantine, forensic preservation, and defensive recovery capacity.
Environmental
Physical or environmental exposure that cannot be honestly represented as a dollar amount alone, including required mitigation and recovery constraints.
Resource
Compute, energy, inventory, labor, capacity, rate limits, scarce resources, or other conserved inputs a machine action can consume.
Contractual
Committed service levels, counterpart obligations, warranties, indemnities, delivery constraints, remedies, and bilateral duties.
Irreversibility
The extent to which an action cannot be unwound. StableMind does not discount irreversibility because another dimension appears well-covered.
Why it matters
04 / CONSEQUENTIAL AI
Machine agency becomes institutional when an action can create obligations faster than a human can notice them.
Reserve value before value moves.
A machine-initiated payment can require exact financial reserve, destination and amount bounds, settlement obligations, and independent reconciliation. Processor acceptance is not settlement, and unknown consequence can keep reserve locked.
Reserve recovery before production mutation.
A technically reversible change may still require rollback capacity, redundancy, blast-radius controls, interruption authority, and an accountable recovery owner before the Action Permit becomes executable.
Reserve for business-state consequences.
Identity, procurement, finance, HR, CRM, or data actions can create contractual, privacy, access, and human consequences even when the underlying API call is ordinary.
Keep consequence ownership local and conserved.
Authority may cross an organizational corridor only inside governed intersections. Each enterprise retains separately owned reserve and proof duties; capacity is not silently netted across parties.
Across the authority chain
Consequence Capital is not a seventh StableMind system. It is a constraint that follows the action.
That is why its lineage touches several systems without becoming owned by the interface that happens to display it.
01Authority Cloudbounds delegated consequence capacity
02ActionGaterequires sufficient current reserve conditions
03Execution Fabricexecutes only inside permit-bound consequence limits
04Guardiancan freeze, narrow, suspend, or contain exposure
05Evidencesupports independent verification and release gates
06Consequence Capitalconserved capacity across the action lifecycle
Constitutional boundaries
05 / CAPITAL IS NOT AUTHORITY
Having capacity to absorb a consequence never means the machine is entitled to create it.
Reserve coverage does not grant authority.
A fully funded financial reserve, available rollback, or strong recovery posture cannot substitute for delegated authority or cause ActionGate to approve an otherwise unauthorized action.
Insurance is not Consequence Capital.
Coverage can supplement recovery after independently proven loss, but exclusions, claims conditions, and future reimbursement do not replace pre-action reserve or widen machine authority.
Modeled safety cannot waive reserve.
A forecast, digital twin, confidence score, historical success rate, or favorable simulation may inform a conservative reserve decision. It cannot erase required capacity or substitute prediction for independent proof.
Human consequence is not monetized away.
StableMind keeps human consequences in their own dimension. Money in another ledger does not make harm acceptable or convert a prohibited action into an authorized one.
Irreversibility cannot be discounted.
An action that cannot be cleanly unwound remains materially different from one with a validated rollback path, even when expected value looks attractive.
The interface cannot release reserve.
A green dashboard, proof summary, paid invoice, customer acceptance, or operator click cannot independently rewrite the governed reserve state. Release follows exact policy and evidence gates.
For evaluators
06 / ASK BETTER QUESTIONS
Questions to ask before an AI agent can create real-world exposure.
- What is Consequence Capital?
- It is explicitly bounded capacity reserved before consequential machine action to absorb, contain, recover from, compensate for, or settle the exposure that action can create. The exact form depends on the consequence domain.
- Is Consequence Capital just money?
- No. Financial reserve is one dimension. StableMind also preserves human, legal and regulatory, operational, data and privacy, security, environmental, resource, contractual, and irreversibility dimensions without pretending they are freely interchangeable.
- Why reserve capacity before an AI agent acts?
- Because consequential automation can create exposure at machine speed. Pre-action reserve turns preparedness into an explicit authorization condition instead of hoping recovery resources exist after the event.
- Does enough reserve mean ActionGate should approve?
- No. Reserve is necessary only when policy requires it, and it is never sufficient by itself. The request must still possess valid delegated authority and satisfy every other deterministic policy, approval, separation-of-duties, revocation, and execution condition.
- When is consequence reserve released?
- Only when the applicable governed release gates are satisfied. Depending on the workflow, that can require independent Proof of Consequence, settlement evidence, resolution of disputes, recovery completion, or other exact conditions. The public website cannot release reserve.
- Can insurance replace consequence reserve?
- No. Insurance may provide additional recovery resources under a separate contract, but it does not create authority, eliminate exclusions, guarantee immediate liquidity, or replace the pre-action capacity required by the machine-authority policy.
Test the boundary
Freeze consequence capacity and watch the authorization state change.
The Authority Lab ↗ makes the reserve condition interactive while keeping every result synthetic and browser-local.
Public truth boundary
This page explains the Consequence Capital architecture. It does not claim a live reserve.
The reserve example is synthetic. StableMind does not claim that this public site holds customer funds, operates a bank account, extends credit, issues insurance, controls a live customer reserve, has settled a customer consequence, or has established realized economic value. Where financial examples appear, they demonstrate the architecture and constitutional boundary rather than a regulated financial service claim.
content_state=ARCHITECTURE_AND_PRODUCT_INFORMATION_WITH_SYNTHETIC_CONSEQUENCE_RESERVElive_customer_reserve_claims=0reserve_release_effects=0financial_service_claims=0website_session_creates_authority=falsepublic_signup_active=false
The customer path
Start with the action you refuse to automate because the consequence matters.
ActionGate remains the commercial front door. In the eventual onboarding flow, StableMind will help a customer make authority and consequence conditions explicit before governed execution. SIO10 explains that model publicly; commercial account bootstrap is introduced in SIO19 and the first governed-action experience remains SIO21.
website_session_creates_authority=false · reserve_release_effects=0 · public_signup_active=false