Claims on this page point to a receipt, a verifier, or a signed artifact; independent confirmation of production receipts is not yet available. We don't ask you to trust us. We give you the tools to check.
Verification is the mechanism that is designed to render the independent verdict on H33 artifacts (not yet available for production receipts). This page describes how that verdict is produced and how you can reproduce it yourself. Why this page exists: security claims are only as good as your ability to check them; independent offline re-verification of production receipts is not yet available.
H33 is a cryptographic substrate that turns every decision — accepted, rejected, hostile — into a portable receipt. The H33-74 primitive produces that portable receipt; Verification renders the verdict on it. Three independent post-quantum signature families are designed to sign every artifact: ML-DSA-65, FALCON-512, and SLH-DSA-SHA2-128f. H33 is designed to bind evidence using three independent post-quantum signature families: ML-DSA, FALCON and SLH-DSA. Full three-family production verification remains pending completion and independent qualification of the FALCON repair. A break in one does not break the receipt. The verifier is open. The receipt format is open. The conformance corpus is open. The story works whether or not H33 the company continues to exist.
That is the load-bearing claim. Everything else on this page is consequences and details.
See also — canonical owners: Verification renders the verdict · H33-74 produces the portable evidence · HATS records continuous attestation · Agent-008 governs decisions · FHE computes on encrypted data · ZK proves without revealing.
Every API decision emits a signed receipt. Accepted requests get a receipt. Rejected requests get a receipt. Hostile requests get a receipt. The signature binds the decision to the policy version, identity, time, and inputs.
Three signature families by design (not yet available in production): lattice (ML-DSA), NTRU lattice (FALCON), hash-based (SLH-DSA). The receipt verifies cleanly under any one of the three. Two breaks would be needed to compromise.
The verifier is open; verification of production receipts is not yet available.
Any change to a signed artifact invalidates the signature. Verification renders a VERIFIED or NOT VERIFIED verdict with the failure mode named. No partial trust.
A receipt-bound decision can be replayed against the same inputs and policy version and reproduce the same outcome. Agent-008 governs which decisions execute inside the authority boundary; Verification renders the verdict when the receipt is replayed. Disputes collapse to facts.
HATS records a continuous attestation stream of control state. Replaces annual questionnaires and sampled audit with verifiable, time-stamped, receipt-bound evidence.
| Framework | Status |
|---|---|
| SOC 2 Type II | Report issued · Drata-managed evidence collection |
| ISO 27001:2022 | In audit · 14-domain Annex A mapping. Details |
| HIPAA | Architecturally aligned via FHE-based encrypted processing (FHE computes on encrypted data). BAA available on request |
| PCI DSS | Architecturally aligned. Details |
| GDPR | DPA available. EU-data residency on Sovereign tier |
| NIST FIPS 203/204/205 | ML-KEM-768, ML-DSA-65, SLH-DSA aligned. Details |
| FedRAMP | Ready for sponsoring agency |
| FedNow | Receipt-bound transaction support. Details |
The test that distinguishes H33's security from competitor security claims is whether a claim can be checked without trusting us. For each claim, here is how to check it, and where independent checking is not yet available.
x-h33-substrate, x-h33-receipt, x-h33-algorithms. The receipt is there whether you got 200 or 401.x-h33-algorithms header lists them by name on responses that carry a receipt; a response may have none.Honest threat modeling matters more than aspirational language. Here's what the substrate protects, what it bounds, and what it does not protect against.
Receipts emitted on every decision. Independent verification and three-family signing are not yet available in production. Hit the API or talk to us.
NIST-standardized post-quantum algorithms — ML-KEM (Kyber, FIPS 203) for key encapsulation and ML-DSA (Dilithium, FIPS 204) for signatures — alongside AES-256-GCM for symmetric encryption, plus FHE and zero-knowledge proofs for computing over encrypted data. KMS keys rotate automatically each year.
Sensitive fields are encrypted at origin with field-level and fully homomorphic encryption. FHE computes on encrypted data, so H33 can process and match without exposing plaintext on shared infrastructure.
H33 has received its SOC 2 Type II report from Prescient Assurance (Security Trust Services Category; April 1 \u2013 July 1, 2026), with continuous monitoring through Drata, and is pursuing ISO 27001:2022. Evidence is collected continuously against 114+ controls, and HATS records a continuous, receipt-bound attestation stream of control state. See the ISO 27001 and SOC 2 pages for detail.
Verification is designed to render the independent verdict (offline verification with the open Verifier CLI is not yet available for production receipts). The evidence they verify is the portable receipt H33-74 produces. For formal procurement review, request the Security Exhibit.