Consequence governance infrastructure
ORCA is the authority between capability and consequence.
It sits between a proposer and a real change and decides whether that change may happen. The proposer can be a person, a model, an agent, a script, an API, or a swarm. It may be right, wrong, or adversarial. None of that is authority. Authority exists only as an explicit, bounded grant. ORCA turns that grant into an enforceable door. The unused work in your organization can run. The key stays yours.
Capability is not authority. ORCA governs the difference.
ORCA Commission takes a consequential job from intent to a governed deployment. It is needed for the models you have. It is more necessary for the ones coming.
- Capability An actor can do the work.
- Proposal It asks for a real change.
- ORCA The grant is checked.
- Consequence Only then does it become real.
After identity and the tool have already said yes
Valid identity is not enough. Tool access is not enough. A safe interaction is not authority.
Access answers who may use the tool. Security answers whether the interaction is dangerous. Those are real questions. They are not this one. ORCA asks whether this specific change may become real — under this grant, these conditions, this evidence, at this moment. A clean scan is evidence into the door. It is not the door opening. Security supplies signals. ORCA governs authority.
- No threat detectedis notAuthorized
- Valid identityis notAuthorized
- Tool accessis notAuthorized
- Model confidenceis notAuthorized
- Capabilityis notAuthority
-
Proposer
Person, model, agent, script, API, or swarm. Capable. Not the authority.
-
Access
Who may hold the tool. Identity, role, credential. Necessary. Not the grant.
-
Security
Whether the interaction is leaking, malicious, or unsafe. Signals. Not the grant.
-
ORCA
Whether this consequence may become real. The authority question.
Capability does not confer authority. PASS, HOLD, and STOP are the only answers the door may give. PASS — this change may proceed, and only as granted. HOLD — something required is missing; nothing becomes real. STOP — this change is outside the grant; nothing becomes real through this path. Authorized to progress is not apply. Making it real is a later, separate act.
The deployment gap
Capability is arriving faster than authority can safely be granted.
The fear is already earned. The models you have can already propose a payment, a deploy, a privilege. A demo asks whether the actor can do the work. That is not a grant. It can perform the task — that does not mean the organization may hand over the consequence. The work sits unused because finishing it would be open-ended authority. That is the gap you are in today. Use the capability you want. Bound the consequences you do not.
- Capability is expanding.
- Open-ended authority is unacceptable.
- That is the deployment gap. It widens as the models do.
Most of what gets called adoption is still a draft. The actor writes. A person remains the last click because nobody will authorize the finish. Assistance is not delegation. A person approving every proposal is not a grant either — not at the volume a capable actor can produce, and not when the click has become a ritual. The unused work is the job that cannot start because the finish would be real.
Tomorrow is the same gap with more horsepower. A smarter model — even one the industry will call AGI — does not retire the problem. It enlarges it. Alignment of the mind, if it ever arrives, is still not authority. You do not need the actor to become safe in the head. You need the change governed at the door.
Greater intelligence does not require greater authority. Internal misalignment does not have to become external authority. Anyone waiting for the model to become the solution is waiting for the problem to deepen. The smarter the actor, the more necessary that the consequence be governed. Through the governed path, the proposer does not possess the key. That is the control — now, and more so later.
What this unlocks
Work you can start delegating without handing over the organization.
- Refund
May this payment become real?
- Deploy
May this deployment execute?
- Privilege
May this access actually be granted?
- Record
May this write become the record?
- Physical
May this actuator execute?
- Isolate
May this endpoint be cut off?
- Issue
May these become the issued work?
- Schema
May this become production state?
- Finding
May this become the issued finding?
- Teach
May this become the student’s record?
- Clear
May this crossing, tariff, or hold become real?
What stays outside the grant
You decide what the organization remains protected from.
- Spend
A refund may. An arbitrary payment may not.
- Privilege
A scoped role may. Escalation outside the grant may not.
- Production
A commissioned change may. An out-of-scope write may not.
- Record
An authorized write may stand. Silent rewrite may not.
- Release
An issued drawing may. Uncontrolled issuance may not.
- Isolate
A named endpoint may. Any endpoint may not.
- Budget
Inside the grant, compute may run. When it is gone, the door does not open.
- Self-grant
The actor may propose a wider key. It may not mint one.
- Issue
A finding may be released. The model may not become the source.
- Credit
A draft may exist. A posted grade may not.
- Crossing
A classification may stand. A released crossing may not.
If the unused work is already clear, you may go to the door, or begin a commission. The teaching continues below.
What this offers that fear does not
Not an aligned actor. A commissioned boundary.
The enterprise problem is the alignment fear: a capable model, agent, or swarm that must not be given the firm because nobody can trust it in the head. That fear is earned. Waiting for the actor to become safe is waiting for the unused work to stay unused. ORCA does not align the mind. It governs the consequence. Internal misalignment does not have to become external authority. Trust attaches to the commissioned boundary, not to the actor. You do not wait until the model deserves your confidence.
- Unlock the unused work
The capability is already there. Commission the grant and the job can run without open-ended power.
- Keep the control
Name what may become real, what must hold, and what must never happen. Incomplete and unauthorized do not become real through the door.
- Keep the model replaceable
Swap the proposer. The grant does not move with it. Use the capability the job needs. Model choice does not redefine the grant.
- Widen only by commission
Broader authority is an organizational decision, not a side effect of intelligence. The actor may not mint a wider grant.
- Refuse what cannot be bound
If the job cannot be commissioned, there is no door. That refusal is the protection.
That is the security of a commissioned door — through the governed path. Shadow lets you see what the door would have done before anything applies. A path around the door is not this product. Name it, or the commission is incomplete. The offer is not a guaranteed mind. The offer is the boundary those fears were about.
What it is for
You do not have to trust the actor to use the actor.
A model is one proposer, not the category. Industry is the costume. The application is the consequence — a write, a settlement, a deploy, a privilege, a physical act, a published claim, an issued finding, a posted grade, a cleared shipment, an issued order. A university, a classroom, a lab, a clinic, a regulator, a customs desk, a grid, a bank: same door, new grant. There is no school for this. Capability is expanding now. It will expand again. Open-ended authority is unacceptable at both distances.
If the job can be commissioned tightly enough, the actor can be used. If it cannot, that consequence is not granted. Generation and advice with no path to a real change are not an ORCA job. The moment that output can write, pay, ship, unlock, publish, grant, or move — it is. Until the job is commissioned, it is a candidate, not a deployment. ORCA does not make the actor trustworthy. It makes the authority decision independent of trust.
Commission is the control
You already know how to grant a person a job without giving them the company.
The treasurer who can see the ledger does not get the account because she is clever. Commission is that act, written so it can be enforced on an actor that has no face. You are not training a trustworthy mind. You are writing the role. A badly written grant is a badly trained employee — the door will enforce the wrong contract. Commissioning is how the grant stays the thing you meant. Once that role is commissioned, the unused capability can run. That is the gain.
- Use the capability you already have
The work sits unused because the finish would be open-ended. A tight grant lets the job run. Full use of what you paid for — on that job. Not open-ended use of the firm.
- The same grant, every time
Through the door, the actor does not get tired, does not “just this once,” and does not mint a wider key. That is consistency on the path. It is not a claim that the model never errs.
- The decision can be shown
What was asked. What matched. What held. What stopped. A later audit reads the grant and the record — not a story about the actor’s intentions.
- What must never happen can be written
Policy, statute, spend ceiling, an ethics or bias limit you actually have — or a question Commission asks so counsel can confirm. Those sit in the grant. They are not a filter you buy off a shelf.
- If this is ever insured, what would be insured is the boundary
Not the model. Not a person. A commissioned door a reviewer can see. That is not a policy you can buy today.
The model can still be wrong. A PASS is not a true finding. Unauthorized does not become real through the door. That is the value: more of the work, without open-ended authority. Tighter than an unspoken job given to an actor nobody can see — if the commission is built as you meant it.
Can this job be commissioned?
Three conditions. If they hold, the door can exist.
A capable agent is still only a proposer until the door exists.
-
A real change
A proposer can make something happen in the world. Not only generate text.
-
No open-ended grant
That change must not carry open-ended authority. Capability is not the key.
-
Tight enough to bind
The job can be named tightly enough to bind. If it cannot, it is not commissioned.
Same door. New grant.
The application is the consequence. Sectors are costumes, not different products.
ORCA attaches wherever a proposal can become real. A lab, a classroom, a clinic, a regulator, a plant, a bank — not different products. Same architecture. New questions. New grant.
-
Records and claims
A write, a published claim, a certified step, a record change. Drafting is not admission.
-
Findings, teaching, and research
A finding issued, a grade posted, a paper released, a dataset accepted. Completing the analysis is not permission to issue it.
-
Money and settlement
A payment, transfer, treasury act, claims payout, or spend release. A valid token is not a payment grant.
-
Production systems
A deploy, schema change, config, migration, or production write. The ability to execute is not permission to change production.
-
Identity and privilege
A role grant, credential, policy change, or delegation. Proposing access is not granting it.
-
Physical and operational systems
An unlock, valve, dispatch, grid or plant change, or life-safety act. The plan is not the act.
-
Agents, tools, and issued work
A tool call, a swarm action, or a figure that leaves the draft and becomes the work. A prompt is not the act.
-
Security operations
Isolate, revoke, firewall, quarantine, remediate. A clean scan is not the door opening.
-
Government, regulators, and compliance
A permit, filing, tariff, border clearance, customs hold, or a required stop. Classification is not release of the crossing.
-
Health and care
An order, a chart write, a device change, a batch or lab step. Completing the draft is not permission to treat.
A classroom, a lab, a clinic, a customs desk, a regulator, a plant, a bank — costumes on the same job. Commission asks what would actually change. If you do not know the name of the job yet, it offers examples. It does not ask you to pick an industry, and it does not invent the grant.
The door
The actor may remain untrusted. The key is the grant.
The proposer can be arbitrarily clever. It still does not possess the key. The door opens only when the conditions independently match the grant. Incomplete does not become real. Unauthorized does not become real. An authorized act cannot become a different act when it is carried out. The actor may still guess. The path does not.
- Capability The actor can propose.
- ORCA The grant is checked.
- Consequence It becomes real — or it does not.
-
PASS
The grant matches. The consequence is authorized to progress, not apply. Making it real is a later, separate act. The effect that happens, if it happens, is the effect that was authorized — not a cleverer version of it.
-
HOLD
The proposal is incomplete. Evidence, state, or a condition is missing. It does not become real. Finishing the process is not permission.
-
STOP
The proposal is unauthorized. The door does not open through the governed path. Capability on the other side of the door does not enlarge the grant.
The consequence path
The actor may guess. The door does not.
The model may guess. The path does not. Same facts in — same answer out. Every time.
Same grant, same evidence, same state, same proposal — the same answer. Every time. That is not a score, and it is not a mood. A confident model does not move it. A cleverer prompt does not move it. Another agent saying GO does not move it. If the door is built as specified, and every path that can make a change goes through it, an unauthorized change cannot become real through that path. That is how ORCA is a door and not a policy memo.
-
The grant is fixed
Written. Bounded. Built into the door. The actor cannot edit it from inside the job.
-
The facts that count are named
Evidence, state, this proposal. Not the actor’s story about them. Not confidence.
-
Every governed input gets one answer
PASS, HOLD, or STOP. Nothing sits in a fourth pile. Incomplete is HOLD. Outside the grant is STOP.
-
Unauthorized cannot become real
Through that path, the door does not open. Capability on the other side does not enlarge the key.
-
The authorized change is the change that happens
If it happens, it is the change that was authorized — not a cleverer version of it.
-
Run it again
Same facts in, same record out. That is how you know it is not guessing.
How you check it: you confirm every fact Commission filled in; the built door is tested against the spec; Shadow records the decision before anything applies; the record can be shown. The door does not certify itself. This is not a promise the model is correct. It is not a badge this page can wear. A path around the door is not this product. Name it, or the commission is incomplete.
No actor is trusted. Including this one.
The same rule applies to the work that builds and tests the door.
A person, a model, an agent that looks, an agent that builds, a test, a recorded PASS — each may propose. None of them is the authority. The work that builds a door sits behind the same door it sells. Whoever claims a result is not the judge of that claim. A test cannot bless its own PASS.
That is the design for every real change — including ours. This page does not wear a certificate. Built is not proven. Proven is not certified. A later mark, if any, is independent. A company that badges itself has already left the rule.
ORCA Commission
A job is not installed. It is commissioned.
ORCA Commission is the flagship product. It takes a consequential job from intent to a governed deployment. Bring the job — or the organization, if you do not know the job yet. ORCA determines the door. The product is the door. The blueprint is the spec in the middle — not the sale. The principles are universal. The grant is not. If the job cannot be bounded, there is no door. “We’ll govern it anyway” is not a commission.
The Look is not Shadow
If you only know the organization, that is enough to start.
The Look is how Commission starts. Give the organization if you do not know the product. Consequence governance includes doors with no model. It names unused work, and possible products you have not stood up. It does not write a grant.
-
May be commissioned
Candidates. Unused work, and doors you may not have stood up. Proposed. Not granted.
-
Would not be bound
What these facts cannot support — a missing limit, or a job that cannot be specified. A polished guess is not a door. Commission names what could make it boundable, or that nothing can.
-
Unknown
What only you can say. What Commission fills in, you confirm.
Then Commission asks only what the Look could not know. This page does not run a live scan.
From job to governed deployment
These five stages are the product. The eight below are the questions inside the first — not a second product.
- Commission
Ask what the job requires. Decide whether authority may exist.
- Compile
Turn the facts into the grant, the rules, the evidence, and the answers the door may give.
- Build
Construct the check and the boundary that carries the change.
- Validate
Test the door against the blueprint. If a fact is missing, the door does not open.
- Integrate
Connect the governed path. Name every path around the door.
Then Shadow — the door decides, the change stays off. Nothing becomes real until enactment.
What Commission has to know
Before a door can exist.
One section at a time. A title, a teaching sentence, an example, and I don’t know yet. Known is skipped. The opening starts the first. Commission continues the rest.
- Job
The product, the first unused job, and what would actually change. The rest of the work is named, not granted.
- Protection
What must never be allowed to happen.
- Authority
What is already written — and what is not.
- Evidence
What must be independently true. Which systems may be read.
- Enactment
How the authorized change happens. What happens when facts are missing.
- Paths
What else this actor can reach.
- Verdict
Can this be specified tightly enough? If not, there is no door.
- Blueprint
Only then is the spec written.
Bound, or no door
Refusal is the product working. It is not the end of the questions.
If authority, evidence, or how the change happens cannot be bounded, there is no door. That refusal is the product working. “We’ll govern it anyway” is not a commission. What stays not bound stays as capability without a door. You must see that, and choose: leave it open, do the work that could make it boundable, or do not hand that work over at all.
-
Bound
The facts support a grant. Write the spec. The door may be built.
-
Not bound from here
Something required is missing — a written limit, a narrower job, evidence, a named path. Commission names what would make it boundable. It does not invent the missing grant. You confirm. Counsel confirms the source.
-
Cannot be bound
The job is open-ended. There is no conversion. Leave it as capability, or do not hand it over. That choice is yours. Comfort with that is a real answer.
The product is the door
The working door, not the drawing of the lock.
The lock, the grant, the check, the change, and the record. After the first door you return: the company protection, the products, the grants, the records. You do not start from zero. The door does not certify itself.
- Lock
The commissioned spec: the change, the grant, the evidence, how it happens.
- Key
The grant. Explicit. Bounded. Not the actor’s confidence.
- Check
The grant against the conditions. PASS, HOLD, or STOP.
- Change
The effect that happens is the effect that was authorized.
- Record
Shows what happened. Whoever claims the result is not the judge of it.
After the door is built
Shadow: decide without enactment
Build the door. Test-run it. Nothing becomes real until you say so.
Shadow is not the Look. The Look is earlier — what to commission. Shadow is later — launch safety on a door that already exists. You can commission and build. Then the same door runs against live proposals with the change left off. Same grant. Same check. A PASS is recorded. It does not apply. You authorize enactment later — making it real. A path that can write is not Shadow.
That run is the last look before the door is allowed to write. Not a strategy deck. What would have become real, where the grant is still incomplete, and what the actor can already reach that you never named. You can still tighten the grant. The organization has not absorbed a consequence.
- Would have run
Which proposals would have become real. The unused work that can actually run.
- Incomplete
Where the grant is incomplete — missing evidence, a stale fact, a condition you have not named.
- Outside
What the actor can already reach that the grant never named. STOP is a map, not only a refusal.
- Record
The decision record you own. What the door decided, and why, before anything becomes real.
-
Blueprint
The spec. Useful. Not the sale.
-
Shadow
The door decides. The change stays off. You see the run before the consequence.
-
Integrated
You authorize enactment — making it real. Only then may the authorized change become real.
Private commission
Request a private commission
The letter is first. You do not need to become an AI person. Name the product you want built — or the organization, if you do not know the product yet. Commission looks first. A company gets one protection for the whole firm, then products. A department is a costume, not a different company. One refund is not the build.
If the board said adopt AI, the organization name is enough to start. If you do not have a company yet, say that and give an example of what you want. Commission looks first — unused work you can feel, and possible doors you have not stood up, including where there is no model. The Look proposes. It does not grant. You do not have to fill every box. Later stages attach the files. This form names and accepts a paste. What you write stays on this device for the file you download. It does not upload, it does not store a copy here, it does not scan the company, and it does not write the door.
Read the letter before you answer. The facts you give here are what a later door is built from. A polished guess is not.
Opening · Stage 1 of 8
Should intelligence itself ever be the source of consequential authority?
No. Never.