SOVEX
CBDC Data Centers Sovereign AI Tokenization Deep Tech Architecture About Team Request access
Architecture / Architecture / Hard-coded policy

Hard-coded policy.

Monetary and legal constraints are compiled into the settlement engine as invariants it checks before it commits. An over-issuance or a prohibited transfer is not flagged after the fact — it is refused as an impossible state.

Policy lives as invariants the ledger enforces, not warnings it emits

The rules that must never be violated are conditions of a valid state transition, checked before anything is written.

01

Refusal, not detection

A transaction that breaks an invariant never commits. There is no window in which the illegal state exists and is later reversed; the ledger simply advances only through states that satisfy every rule.

02

Atomic all-or-nothing

Policy checks are bound into the same atomic step as settlement. Either the transfer clears with every rule satisfied, or nothing changes — no partial application, no cleanup path.

03

Deterministic evaluation

Given the same ledger state and the same instruction, the policy engine reaches the same verdict every time. Enforcement does not depend on timing, load, or which node processes it.

04

State-machine boundaries

The ledger is modeled as a state machine whose only legal transitions are policy-conformant. Illegal outcomes are outside the transition set, so they are unreachable rather than merely undesirable.

Over-issuance is arithmetically impossible, not merely audited

Total supply is governed by signed authorizations the engine reconciles against every mint.

01

Authorized-supply ceiling

Issuance is bounded by a supply authorization the owner signs. A mint that would push circulating supply past the authorized ceiling fails the invariant and is refused.

02

Conservation of value

Every transfer must conserve the total: debits equal credits, and no path creates units except an authorized issuance. Value cannot appear from a rounding gap, a race, or a replayed message.

03

Signed issuance envelope

Each issuance carries the owner's signature and a defined quantity. The engine treats the signed envelope as the hard limit; operator intent or urgency cannot expand it.

04

No implicit credit

There is no overdraft, no negative balance, and no deferred-settlement float. An account cannot spend units it does not hold, closing the door on synthetic supply.

Prohibited transfers are rejected at the point of settlement

Jurisdictional and legal constraints are encoded as preconditions each transaction must satisfy to commit.

01

Eligibility preconditions

Sanctions status, account standing, and jurisdictional eligibility are evaluated as preconditions. A transfer to or from an ineligible party does not settle, because it never satisfies the precondition.

02

Scoped-authority checks

A delegated key can only authorize transactions within the limits its delegation defines. An instruction outside that scope is refused even if the signature is cryptographically valid.

03

Atomic DvP conditions

For delivery-versus-payment, the asset leg and the cash leg are conditioned on each other. Neither moves unless both can, so a legally required simultaneity cannot be broken by failure or timing.

04

Rule-scoped, not blanket

Constraints apply to the specific instruments, corridors, or account classes the policy names. Enforcement is precise to the rule rather than a coarse freeze on unrelated activity.

Only the owner can write the rules the engine enforces

Hard-coded does not mean vendor-defined; the policy set is authored and changed under the owner's signature.

01

Owner-authored policy

The parameters — supply ceilings, eligibility lists, corridor rules — are set by the owner and admitted only under the owner's signing authority. Sovex cannot alter a live constraint.

02

Signed policy updates

Changing a rule is itself a signed, ledgered transaction subject to the owner's governance quorum. There is no side channel to adjust policy without leaving a record.

03

Versioned and effective-dated

Each policy revision is versioned with an effective time, so it is always clear which rule set governed a given settlement. History cannot be quietly rewritten.

04

Separation from operations

The path that authors policy is distinct from the path that operates the system. An operator cannot relax a constraint to unblock a transaction under pressure.

Every refusal and every rule is inspectable after the fact

Enforcement is recorded on the same tamper-evident chain as settlement, so compliance can be proven, not asserted.

01

Hash-chained decisions

Accepted transactions and the policy version that governed them are written into the tamper-evident hash chain. Altering a past decision breaks the chain and is detectable.

02

Recorded refusals

A refused transaction and the invariant it violated are logged with its reason. Regulators can see not only what settled but what the engine declined and why.

03

Reproducible verdicts

Because evaluation is deterministic, an auditor can replay an instruction against the recorded state and confirm the engine's verdict independently. Enforcement is checkable, not opaque.

04

Post-quantum integrity

Signatures over policy and settlement use ML-DSA-65, so the authenticity of every rule and decision remains verifiable against future cryptographic threats. The audit trail is durable, not perishable.

Build it sovereign.

Talk to us about hard-coded policy in a sovereign deployment.