Rules-as-code that can be proven.
QuantZK lets your software prove it followed a rule set, without exposing private data. Verification works offline, forever, without our servers.
npm install @quantzk/attest
pip install quantzk
go get github.com/quantzk/vdi-verifier-go
# coming soon
import { attest, verify } from '@quantzk/attest'; const proof = await attest('fair-lending-v1', { inputs: { credit_score: 720, income: 85000 }, outcome: 'approved' }); const { valid } = await verify(proof); // true
Three lines. The manifest defines the rules. The proof is automatic. Verification is independent.
| What the verifier learns | What stays private |
|---|---|
| Decision followed the manifest's rules | Input data (credit score, income, etc.) |
| All policy constraints were satisfied | Model weights and algorithm |
| Constraints were reviewed by named authority | How compliance was evaluated |
| Proof is mathematically valid (Groth16) | Any internal state |
| Manifest | Regulation | What it proves |
|---|---|---|
fair-lending-v1 |
Fair Lending Act | Lending decision was non-discriminatory |
eu-ai-act-v1 |
EU AI Act | AI system meets EU requirements |
hipaa-v1 |
HIPAA | Healthcare data handled correctly |
soc2-access-v1 |
SOC 2 CC6 | Access controls operate effectively |
nist-ai-rmf-v1 |
NIST AI 100-1 | AI follows risk management framework |
model-safety-v1 |
Internal / NIST 600-1 | AI output satisfies safety constraints |