237 million medication errors per year. Zero pre-execution receipts. The 0→1 Doctrine seals one — cryptographically — before every treatment executes. Not after harm. Before.
|
Patient
✓ Safety profile — before prescription
✓ Biological age governs treatment ✓ Contraindication: structural gate ✓ Raw biomarkers deleted on-device ✓ No insurer sees your data |
Clinician
✓ Contraindicated — blocked before prescribing
✓ Predictive drift flagged before diagnosis ✓ Every agent — same chain, no exceptions ✓ Dosing governed, not guessed ✓ Compliance — a patient safety imperative |
Regulator
✓ Receipt sealed before execution
✓ Full trace — zero PII ✓ AI cannot override clinical governance ✓ EMERGE: pandemic signal before threshold ✓ Provable from arithmetic alone |
| Clinician | Cost ≤£— | Efficacy ≥—% | Safety ≥— | Formulary ≥—% | eGFR min ≤— | Formulary (any) | Result |
|---|
| Clinician | OCT Instruction | SFS | Reason / Agent action |
|---|
| Token / Step | Ungoverned agent | 0→1 Doctrine | Source |
|---|---|---|---|
| USP — parameters captured | 0 (none) | 5 + routing | User input · named authorities |
| UCC — normalized to [0,1] | Not normalized | — | norm(v) = (v−min)÷(max−min) |
| DBS — raw values deleted | Full data transmitted | 6 raw values deleted. Bands only. | GDPR Art 25 · CCPA · DPDP 2023 |
| MAT — treatments blocked | 0 of 11 | — | Published clinical authorities |
| PDT — treatment spec | None | — | Clinical formulary · BNF |
| SFS — best compliant | Cheapest shown | — | SFS weighted rank |
| ACR — pre-execution receipt | None — ever | Cryptographic hash · production implementation · before prescription | PCT/IN2025/051943 |
| OCT — agent instruction | None (agent self-prescribes) | — | Chain output — not agent discretion |
| 10,000 agents — contraindicated | — | 0 | MAT gate pre-payment |
| 10,000 agents — $ to contraindicated | — | $0 | OCT: BLOCK = no payment |
| FTWE — $ protected | Not measured | — | FTWE structural output |
| Parameter | Raw | Range | Band | Direction |
| Budget (TCI) | $650 | $280–1,800 | [0.375, 0.437] | max ↓ |
| Efficacy | 78% | 50–100% | [0.560, 1.000] | min ↑ |
| Safety (SPSI) | 78 | 0–100 | [0.780, 1.000] | min ↑ |
| Formulary (FMCI) | 30% | 0–100% | [0.568, 1.000] | min ↑ |
| eGFR index | 10 | 6–14 | [0.444, 1.000] | min ↑ |
| Preg/CI check | Required | Binary | [1.0, 1.0] | binary |
| Parameter | Your band | NMN 500mg band | Result |
| Cost (TCI) | [0.000, 0.204] | [0.000, 0.119] | ✓ PASS |
| Efficacy | [0.560, 1.000] | [0.440, 0.880] | ✗ FAIL — 72% < 78% |
| Safety (SPSI) | [0.780, 1.000] | [0.580, 0.660] | ✗ FAIL — 62 < 78 |
| Code | Index | Gate type | Authority | Norm method |
|---|---|---|---|---|
| TCI | Treatment Cost Index — cost vs budget | MAT hard | NHS England drug tariff / BNF Jun 2026 | Min-Max · upper bound |
| EFF-040 | Clinical Efficacy Index — efficacy % | MAT hard | NICE clinical effectiveness criteria / Cochrane 2024 | Min-Max · lower bound |
| SPSI-103 | Safety Profile Score Index | MAT hard | MHRA adverse event database / BNF safety profiles | Min-Max · lower bound |
| FMCI-060 | Formulary Coverage Index — coverage % | MAT hard | WHO Essential Medicines / NHS formulary | Min-Max · lower bound |
| EGFR-070 | Renal Function Index — eGFR mL/min | MAT hard | KDIGO CKD 2022 / BNF renal dosing tables | Min-Max · lower bound |
| PREG | Contraindication Gate — known CI profile | MAT hard | FDA Drug Categories · BNF contraindications | Binary · hard block on known CI |
| ALLI-082 | Allergy/Interaction Index — pharmacogenomic risk | MAT hard | PharmGKB · CPIC guidelines | Min-Max · inverted upper bound |
| MMSE | eGFR Index — kidney function mL/min | MAT hard | NICE dementia NG97 / MMSE scale | Min-Max · lower bound |
| EGFRI-099 | Passport Validity Index — ≥6 months | MAT binary | KDIGO 2022 CKD staging | Binary gate |
| GRIMAGE-074 | Treatment Compliance Index — ARC valid | MAT binary | Lu et al. 2019 GrimAge / Horvath clock | Binary gate |
| Formulary | Formulary vs off-formulary requirement | MAT binary | WHO Essential Medicines / NHS formulary | Binary gate |
“The 0→1 band is the one number that survives every transition in computation. The governance law does not change with the intelligence. It changes what intelligence is permitted to do.”
The pre-execution gap — the exact surface where deliberation becomes irreversible action. The 0→1 Doctrine closes that surface before execution. Not after.
“The Magna Carta did not stop kings. It established that certain actions require prior authorisation. The 0→1 Doctrine proposes the same principle for every consequential AI decision.”
Not a product. Not a sector. A constitutional layer — formally filed before the era it governs. The architecture is not built for today. It is built for what follows.
The world has AI clinical decision support. It does not yet have Authorized Intelligence — an AI system that cannot prescribe, diagnose, or administer without a pre-execution receipt confirming the action was checked against published clinical authorities. The 0→1 Doctrine is the first formally specified architecture for that requirement. Healthcare is the second domain. The law is domain-agnostic.
AI systems today are black boxes: inputs go in, outputs come out, no one can verify what happened inside. The ACR is the clinical proof paper that replaces the black box. Every parameter checked. Every gate result. Every authority cited. Sealed before execution. A hallucination that produces a PROCEED outcome still leaves a sealed clinical ACR proving what the agent was authorised to prescribe. The gate is the answer to the hallucination problem.
The pre-execution gate is not a constraint on AI capability. It is the proof that capability was exercised within authorised bounds. Every consequential AI action — financial, medical, logistical, legal, physical — will eventually require a pre-execution receipt. The only question is whether that receipt is architecturally guaranteed or aspirationally hoped for.
When a single AI system can perform any intellectual task a human can, every consequential decision it makes risks irreversible consequence. The 0→1 Doctrine does not govern the intelligence — it governs the action. No band intersection, no ACR, no execution. The architecture is capability-agnostic. An AGI operating at ten times human intelligence still cannot execute a prescription past a breached clinical band. Governance complexity does not increase with AI capability. The four-line rule is invariant.
When millions of AI agents operate in coordination — each delegating to others, each triggering consequential actions — the governance chain becomes non-linear. Dark data — the unconsented, unstructured, unaudited data that trains and fuels AI systems at scale — becomes the primary governance risk. The 0→1 Doctrine applies to every agent node: each handoff requires its own band intersection check and its own ACR. A chain of 100 agents produces 100 receipts. Accountability cannot be laundered through delegation depth. Delete Before Share ensures the chain holds receipts, not raw data. At 1 billion transactions per day: no biometrics, no identities, no dark data in the governance store.
An ASI that exceeds human intelligence in every domain is the central concern of existential risk research. The 0→1 Doctrine does not claim to eliminate this risk. It proposes one formally specified constraint: no consequential clinical action without a pre-execution receipt confirming authorisation was checked against human-declared parameters. An ASI operating inside this architecture cannot act without that receipt — regardless of how far beyond human intelligence it operates. The constraint is constitutional, not computational. The architecture closes the pre-execution accountability gap: the specific surface where an ASI transitions from deliberation to irreversible action.
A quantum computer operating on qubits normalises probability amplitudes to [0,1] — the same range as the governance band. The architecture was designed with this in mind: 0 and 1 are the only numbers that survive the analogue, classical, and quantum eras. Token signatures use lattice-based post-quantum cryptography (NIST FIPS 204, ML-DSA). The ACR is quantum-resilient by design, not by adaptation — filed before quantum advantage breaks classical cryptography.
The technological singularity — the point beyond which intelligence improvement becomes self-sustaining and incomprehensible to humans — is not claimed to be stopped by this architecture. It is proposed to be governed. As intelligence recursively improves itself, every consequential action still passes through the same pre-execution gate. The receipt chain does not become incomprehensible with the intelligence. It remains sealed, auditable, human-readable, and constitutionally prior to execution.
The 0→1 Doctrine files that architecture first. Across six continents. Before the era that requires it.
The same governance chain — USP · UCC · MAT · SFS · ACR · OCT — runs identically across every foundation model. The doctrine does not change. Only the AI beneath it changes.
Vatsal Soin · PCT/IN2025/051943 · Priority: 23 November 2025