Predicate ZK develops zero-knowledge proof infrastructure for regulated industries. Our foundational technology — the Separation Predicate™ — enables any party to prove that a regulated decision was reached without reference to protected data, without disclosing any underlying information to any counterparty. Currently deployed for AI-assisted prior authorisation and 340B Drug Pricing Programme compliance.
Regulators across healthcare, financial services, and AI are mandating demonstrable compliance — proof that a decision was reached correctly, without reference to prohibited inputs. Predicate ZK provides the cryptographic infrastructure that satisfies that mandate, across every regulated sector, from a single foundational proof system.
EU AI Act · Media integrity certification · Content provenance · AI bias obligations
Government & procurement
Federal set-aside compliance · National security data separation · Defence procurement
How it works
One proof. Every regulated decision.
Across every regulated industry, the compliance problem reduces to the same underlying question: can one party prove to another that a decision was reached without reference to a prohibited input — without disclosing the data, the model, or any private information? That is precisely the problem the Separation Predicate™ solves.
The core insight
Every regulated industry contains the same unsolved problem. One party must prove to a counterparty that a decision was not contaminated by a protected or restricted input — without disclosing any underlying data to anyone.
This is a process-independence claim. It is structurally different from a fairness claim, an accuracy claim, or an audit trail. Regulators are increasingly mandating this demonstration. The Separation Predicate™ provides it.
Why regulatory compliance requires this
The EU AI Act, CMS-0057-F, HRSA 340B programme requirements, FinCEN AML obligations, and a growing body of federal and state regulation share a common characteristic: they require organisations to demonstrate that automated decisions comply with specific rules — not merely assert it. Statistical audits conducted after the fact are insufficient. Per-decision proof, generated at the moment of the decision, is what the regulatory direction demands.
Predicate ZK provides that proof layer. The proof is generated automatically at decision time, is 256 bytes, verifiable in under three milliseconds by any authorised party, and requires no disclosure of underlying data, model weights, or proprietary system details to any counterparty.
Platform architecture
Predicate ZK operates as a platform technology and IP entity. Each sector application is an independent operating entity holding an exclusive vertical licence from the platform. The central technical infrastructure — ZK circuit engineering, trusted setup ceremonies, verification key registry — operates at the platform level and serves all verticals.
Zero-knowledge proof engineering is a specialised discipline. Circuit design, trusted setup ceremonies, and verification key management require dedicated infrastructure that sector-specific entities cannot operate independently. The platform provides this as a shared service.
For regulated entities
Compliance officers, legal teams, and regulators receive cryptographic proof of decision independence — per transaction, in real time, without any counterparty needing to access private data. The proof is the compliance record.
Intellectual property
Foundational patent applications filed April 2026 · Separation Predicate™ coined and formally defined · Trademark applications filed · Portfolio covers healthcare, financial services, AI, media, clinical, and government sectors
The Separation Predicate™ applies wherever a regulated decision must be proved independent of a protected input. The proof is generated as a cryptographic sidecar to existing decision infrastructure — additive, not invasive — and produces a single evidence format that any regulator, counterparty, or court can verify independently.
Predicate RX™
Prior Authorisation & 340B Drug Pricing Compliance
Health plans · PBMs · HRSA · CMS · State AI-in-UM regulation
Prior authorisation — primary focus
AI-assisted utilisation management is now subject to specific and tightening evidentiary requirements. State AI-in-UM statutes are in force across more than twenty states. The NAIC AI Model Bulletin has been adopted widely. CMS-0057-F mandates FHIR-based prior authorisation APIs and public metric reporting with effective dates in 2026 and 2027. Delegated-UM counterparties are requiring their delegates to produce AI governance evidence. Plaintiffs’ counsel in ERISA and state bad-faith cases have learned to plead that AI drove adverse determinations, opening discovery into model versions, pipeline composition, and feature independence.
Existing attestation and sampling-based audit approaches answer these questions by assertion. Predicate RX answers them with cryptographic evidence that any counterparty, regulator, or court can verify independently — in milliseconds — from the certificate itself and a published verification key. No access to models, clinical criteria, or protected health information required.
What Predicate RX proves for prior authorisation
Clinical independence
The clinical classifier operated structurally independent of cost features — including under statistical controls for clinical complexity
Model version identity
Each determination is cryptographically bound to the specific model version that produced it — satisfying model inventory and version-control requirements
Pipeline integrity
Multi-step AI pipelines preserved integrity end-to-end — prohibited influence cannot be introduced at an intermediate step and emerge at the output
Tamper-evident history
Decision records are cryptographically chained — any modification to any earlier determination invalidates all subsequent certificates
Episode consistency
Independence maintained across all decisions in a care episode — not just point-in-time but longitudinal
Cross-entity attestation
Delegated-UM counterparties receive cryptographic attestation of AI governance without accessing models, criteria, or member data
One certificate. Every audience.
Today, payers produce different compliance artefacts for CMS, state insurance commissioners, delegated-UM counterparties, internal audit, and litigation discovery — each requiring its own review cycle, attestation chain, and reconstruction work. With Predicate RX, every audience verifies the same 256-byte certificate against the same public verification key. The marginal cost of producing evidence for a new regulator, a new counterparty, or a new discovery category approaches zero.
Deployment model
Predicate RX operates as a cryptographic sidecar to existing utilisation management infrastructure. It does not replace clinical classifiers, decision engines, criteria libraries, or appeals workflows. It attaches at defined integration points and emits a certificate alongside each determination. The architecture is licensed and deployed entirely inside the operator’s own security boundary — Predicate ZK delivers the circuit source, proving and verification implementations, and reference integration kit. No models, clinical criteria, determinations, or protected health information cross the boundary at any point.
Regulatory drivers: CMS-0057-F prior authorisation rules (effective 2026–2027) · State AI-in-UM statutes (20+ states) · NAIC AI Model Bulletin · Gold carding requirements · ERISA and state bad-faith AI litigation · Delegated-UM counterparty audit obligations
340B drug pricing compliance
The 340B Drug Pricing Programme requires manufacturers to provide covered drugs at ceiling prices to qualifying healthcare entities. Proving compliance at the transaction level — without disclosing confidential pricing data or protected health information — has been structurally impossible with conventional methods. Predicate RX provides six proof circuits covering every aspect of 340B verification: ceiling price, patient eligibility, covered entity status, contract pharmacy independence, duplicate discount prevention, and PBM settlement fairness. Each proof is 256 bytes, verifies in under three milliseconds, and requires no disclosure of PHI, pricing data, or claims data to any party.
EU AI Act · Content authenticity · Broadcast · Digital media
AI-generated and AI-modified content is subject to mandatory disclosure requirements under the EU AI Act Article 50, with enforcement beginning August 2026. Predicate MIC provides cryptographic proof that a piece of content was produced by a specific declared process, independently of any prohibited manipulation classifier. Broadcasters, publishers, and platforms can certify content provenance and editorial independence without disclosing proprietary editorial systems or source material.
Enterprise decision systems — hiring, credit, insurance underwriting, medical triage — must demonstrate independence from protected-characteristic classifiers. Predicate AI provides proof circuits for per-decision certification, enabling organisations to meet EU AI Act obligations and anti-discrimination requirements without disclosing model architecture or individual applicant data to any regulatory counterparty.
Regulatory drivers: EU AI Act Article 9 high-risk AI requirements · EEOC algorithmic hiring guidance · CFPB AI lending obligations
In development
Predicate FIN™
AML / BSA Transaction Compliance
Bank Secrecy Act · FinCEN · OFAC · CFTC
AML transaction monitoring systems must demonstrate that risk scoring is independent of protected-characteristic classifiers — a requirement emphasised by FinCEN guidance and fair banking obligations. Predicate FIN provides proof circuits enabling financial institutions to certify monitoring independence without disclosing proprietary scoring models or customer transaction data.
Regulatory drivers: Bank Secrecy Act · FinCEN AML compliance programme requirements · Fair banking obligations · OFAC sanctions screening independence
In development
Predicate CLIN™
Clinical Trial Integrity
FDA · 21 CFR Part 11 · ICH E6 · FDA Diversity Action Plan
Clinical trial eligibility determination and outcome assessment must be provably independent of demographic and financial classifiers. Predicate CLIN provides cryptographic proof of blinding integrity and outcome independence, without disclosing patient data or the sponsor's proprietary trial algorithms.
Regulatory drivers: FDA Diversity Action Plan · 21 CFR Part 11 electronic records · ICH E6 GCP blinding requirements
Forthcoming
Predicate EDC™
Electronic Data Capture
FDA · EMA · 21 CFR Part 11 · HIPAA
A clinical trial data capture platform with separation proofs built into the data architecture at the point of entry. Eligibility and outcome classifiers are separated by construction rather than by protocol, providing a stronger compliance guarantee than audit-trail approaches and satisfying evolving FDA and EMA requirements for demonstrable data integrity.
Regulatory drivers: FDA electronic submissions · EMA clinical data standards · 21 CFR Part 11 audit trail requirements
Forthcoming
Predicate GOV™
Federal Procurement & National Security
FAR · SBA · SDVOB · DBE · National Security
Federal procurement set-aside requirements — small business, service-disabled veteran-owned, disadvantaged business enterprise — require demonstrable compliance without disclosing ownership structures or sensitive financial data. Predicate GOV provides proof circuits for set-aside certification, and extends to national security applications involving dual-stream data separation.
Regulatory drivers: Federal Acquisition Regulation · SBA set-aside compliance · SDVOB verification requirements · National security data separation mandates
A zero-knowledge proof construction that proves two classifiers operating on shared data are computationally independent of each other — without disclosing the data, either classifier, or any intermediate computation to any party. Coined and formally defined by Predicate ZK (Lithosense LLC), April 2026. The foundational circuit for AI-assisted prior authorisation, 340B compliance, and every other regulated application in the portfolio.
Formal Definition — Separation Predicate™
Let D be a decision classifier and P be a protected classifier, both operating over a shared input space X.
The Separation Predicate S(D, P, x) evaluates to true if and only if the computation of D(x) is provably independent of P(x) — the witness used to compute D(x) does not include P(x) or any function of P(x) as an input.
π_S := ZKProof { ∃w : D(x, w) = d ∧ w ∩ f(P(x)) = ∅ }
If D and P are genuinely independent, an honest prover can always produce a valid separation proof.
Soundness
A prover cannot produce a valid separation proof if D received P's output as an input. The proof cannot be forged.
Zero knowledge
The proof reveals nothing about x, D, P, the witness, or the output — beyond the independence claim itself.
Predicate RX™ — six proof circuits
Six proof circuits for 340B Drug Pricing Programme compliance. Each circuit addresses one specific compliance claim, generating a 256-byte non-interactive proof at the moment of the transaction.
ZK-1
Ceiling price verification Proves transaction price ≤ ceiling price without disclosing either value
ZK-2
Covered entity eligibility Certifies 340B programme eligibility without disclosing entity data
ZK-3
Contract pharmacy independence Dispensing classifier independence from patient eligibility classifier
ZK-4
Patient eligibility Certifies patient eligibility without disclosing protected health information
ZK-5
Duplicate discount prevention Proves no duplicate discount without sharing Medicaid claims data
ZK-6
PBM settlement fairness Settlement independence from manufacturer rebate data
Technical reference documents published by Predicate ZK (Lithosense LLC). All documents are timestamped on publication. Document identifiers are citable in regulatory, legal, and academic contexts.
PZK-WP-001April 2026Narrative Overview
The AI Accountability Standard: How Mathematics Solves the Compliance Problem for Artificial Intelligence
Full narrative overview of the Separation Predicate™ framework. Covers the accountability gap in regulated AI decision-making, the mathematical solution, the seven accountability questions, the justice mechanism, sector applications, and the portfolio architecture. Written for a general informed audience.
The Separation Predicate™: A Formal Definition for Zero-Knowledge Compliance Verification
Formal technical definition of the Separation Predicate™ as a zero-knowledge proof construction. Covers the proof architecture, formal statement, three proof properties, and the thirteen foundational platform applications. Establishes coined terminology: separation predicate, separation proof, separation proof circuit.
Predicate RX™: Zero-Knowledge Proof Circuits for 340B Drug Pricing Programme Compliance
Specification for six cryptographic proof circuits (ZK-1 through ZK-6) for 340B compliance verification. Covers ceiling price, patient eligibility, covered entity status, duplicate discount prevention, and PBM settlement fairness. Each circuit produces a 256-byte non-interactive proof. No PHI, pricing, or claims data disclosed.
Public Comment to HRSA: Zero-Knowledge Verification as the Technical Solution to 340B Programme Compliance
Filed in the HRSA notice-and-comment period. Positions Predicate RX™ as the technical mechanism that satisfies the verification requirement HRSA has identified in the 340B programme. Establishes Predicate ZK's presence in the regulatory record.
Predicate MIC™: Cryptographic Content Provenance and Media Integrity Certification
Specification for media integrity proof circuits under EU AI Act Article 50 requirements. Covers content provenance certification, editorial independence proof, and AI-generated content disclosure compliance. Forthcoming.
Forthcoming
PZK-WP-006ForthcomingTechnical Specification
Predicate AI™: Separation Predicate Application to Enterprise AI Decision Certification
Proof circuits for enterprise decision system compliance under EU AI Act, EEOC guidance, and anti-discrimination obligations. Per-decision certification of classifier independence without model disclosure. Forthcoming.
We work with regulated entities, healthcare organisations, financial institutions, regulatory bodies, and technical collaborators. Select the enquiry type that best describes your interest.
Healthcare & pharma
340B & Prior Auth
Manufacturers, covered entities, PBMs, and health plans exploring Predicate RX™ deployment for 340B compliance and prior authorisation certification.
ZK researchers, academic institutions, and technical collaborators interested in the Separation Predicate™ architecture, circuit specifications, or collaboration.