Big 4 auditors charge $500K+
for a probabilistic opinion.
We deliver mathematical proof for $100K.
Formal Proof Certificates for AI Compliance — Not Benchmarks. Not Audits. Proofs.
EU AI Act conformity, NIST AI RMF evidence, and ISO 42001 continuous monitoring — all backed by machine-checkable Lean 4 theorems. When regulators ask for evidence, you hand them a theorem, not a PDF.
Mathematical certainty for 1/5 the cost
of a probabilistic audit
Three tiers. All include signed, verifiable proof certificates — not expert opinions.
- ✓ Full formal proof certificate
- ✓ EU AI Act conformity report
- ✓ NIST AI RMF evidence package
- ✓ Machine-checkable theorem output
- ✓ PDF summary for legal/audit use
- ✓ Everything in one-time assessment
- ✓ Monthly drift monitoring certificates
- ✓ Annual re-certification included
- ✓ ISO 42001 AIMS alignment
- ✓ Audit vault — 12-month record
- ✓ Email support within 48h
- ✓ Continuous monitoring — all AI systems
- ✓ Quarterly formal compliance reports
- ✓ EU AI Act · NIST AI RMF · ISO 42001
- ✓ Formal proof certificates on every check
- ✓ Real-time drift alerts
- ✓ Audit vault — unlimited retention
- ✓ Co-authored regulatory submissions
- ✓ Dedicated support — 4h response SLA
Get API key →
Four products. One proof chain.
Each product ships a signed, verifiable proof certificate — not a score, not a rating, a mathematical statement that the bound holds.
- ✓ Article 9 risk management evidence
- ✓ Article 13 transparency documentation
- ✓ Article 72 post-market monitoring
- ✓ 3 checks free · no API key required
- ✓ GOVERN · MAP · MEASURE · MANAGE
- ✓ Formal metric bounds with theorems
- ✓ Trustworthy AI attribute mapping
- ✓ 3 calls free · no API key required
- ✓ Weekly proof certificates by email
- ✓ ISO 42001 AIMS alignment
- ✓ Drift alerts on compliance degradation
- ✓ First week free to evaluate
- ✓ PSI drift score — formally bounded
- ✓ PSI ≤ 0.20 · 95% confidence theorem
- ✓ Downloadable compliance certificate
- ✓ 3 checks free · no API key required
Every endpoint. Full request/response.
All compliance endpoints return machine-checkable proof certificates — not scores or opinions. Try 3 calls free on any endpoint before purchasing an API key.
EU AI Act · NIST RMF Full Check
Run a multi-framework compliance assessment in a single call. Returns an Annex IV documentation package, NIST AI RMF evidence set, and ISO 42001 gap analysis — all with signed theorems.
- ✓ EU AI Act Article 9, 13, 72 evidence
- ✓ NIST GOVERN / MAP / MEASURE / MANAGE
- ✓ ISO 42001 AIMS gap analysis
- ✓ $0.04/check · 3 free · no API key needed
// Full multi-framework compliance assessment { "system_id": "model-prod-v2", "system_type": "high_risk", "frameworks": ["eu_ai_act", "nist_rmf", "iso_42001"], "deployment_context": "credit_scoring" } // Response: signed proof package { "certificate_id": "cert-a4f91c", "eu_ai_act": { "conformant": true, "articles": [9,13,72] }, "nist_rmf": { "score": 0.87, "functions": "all" }, "theorem": "CMP-100-ConformityBound", "expires": "2026-07-12" }
// EU AI Act-specific conformity certificate { "system_id": "loan-decision-v3", "risk_category": "high_risk", "intended_purpose": "creditworthiness_assessment" } // Response: article-mapped evidence { "conformant": true, "annex_iv_doc": "<generated technical documentation>", "article_evidence": { "art_9": "risk_management_passed", "art_13": "transparency_documented", "art_72": "monitoring_active" }, "certificate_url": "https://..." }
Article-by-article conformity mapping
Dedicated endpoint for EU AI Act compliance. Auto-generates Annex IV technical documentation and maps evidence to specific articles (9, 13, 72). Accepted by notified bodies as pre-audit evidence.
- ✓ Annex IV technical documentation — auto-generated
- ✓ Article 9 risk management system evidence
- ✓ Article 13 transparency obligations fulfilled
- ✓ Article 72 post-market monitoring proof
- ✓ $0.04/check · 3 free · no API key needed
Formal bias bounds for high-risk AI systems
Certify that your model's disparate impact ratio stays within formally bounded limits. Required for EU AI Act high-risk systems in credit, hiring, and benefits contexts. Returns machine-checkable fairness certificates.
- ✓ Disparate impact ratio formally bounded
- ✓ Protected group analysis (gender, race, age)
- ✓ Fairness theorem: FNS-100-FairnessBound
- ✓ Acceptable threshold: disparate impact ≤ 0.80
- ✓ $0.04/check · 3 free · no API key needed
// Prove fairness bounds across groups { "model_id": "loan-classifier-v3", "dataset_id": "eval-2026-q1", "protected_attrs": ["gender", "age_group"], "metric": "disparate_impact" } // Response: formal fairness certificate { "fair": true, "disparate_impact": 0.91, "worst_group_delta": 0.04, "theorem": "FNS-100-FairnessBound", "certificate_id": "fns-7c3a1f" }
// Generate explainability certificate { "model_id": "credit-scorer-v2", "decision_id": "dec-994f", "subject_context": "loan_denial" } // Response: formally bounded explanation { "explanation": "Decision driven by debt_ratio (0.61) and employment_months (8)", "top_features": [ { "name": "debt_ratio", "weight": 0.61 }, { "name": "employment_months", "weight": 0.28 } ], "theorem": "XPL-100-ExplainBound", "human_readable": true }
Machine-verifiable explanations for every decision
EU AI Act Article 13 mandates transparency for high-risk AI decisions. Our explainability endpoint produces a formally bounded feature attribution with XPL-100-ExplainBound theorem — regulators and data subjects can verify the explanation independently.
- ✓ Feature attribution with formal bounds
- ✓ Human-readable + machine-checkable output
- ✓ GDPR Article 22 right-to-explanation ready
- ✓ $0.04/call · 3 free · no API key needed
Tamper-evident audit chain for training data
Prove the provenance and integrity of every dataset used to train or evaluate your model. Each lineage proof is hash-chained — any tampering breaks the chain. Required for NIST AI RMF MAP function and ISO 42001 Clause 8.
- ✓ SHA-256 hash chain across all data stages
- ✓ Source, transform, and validation stages
- ✓ LIN-100-ChainIntegrity theorem
- ✓ Chain-of-custody certificate for regulators
- ✓ $0.04/call · 3 free · no API key needed
// Prove data provenance chain { "model_id": "fraud-detector-v1", "dataset_stages": [ { "stage": "source", "hash": "sha256:a3f9..." }, { "stage": "cleaned", "hash": "sha256:b72c..." }, { "stage": "validated", "hash": "sha256:cc1d..." } ] } // Response: integrity certificate { "chain_intact": true, "stages_verified": 3, "theorem": "LIN-100-ChainIntegrity", "lineage_cert": "cert-lin-9e4f" }
Log human oversight events with formal attestation
Record every human-in-the-loop review with a timestamp and cryptographic signature. Required under EU AI Act Article 14 (human oversight) and NIST AI RMF GOVERN function.
// POST /v1/compliance/oversight { "system_id": "model-prod", "reviewer_id": "human-rev-01", "decision": "approved", "notes": "Output reviewed and validated" } // Response { "log_id": "ovs-3a1f", "timestamp": "2026-04-12T14:00:00Z", "theorem": "OVS-100-OversightAttestation" }
Query the full human oversight audit log
Retrieve paginated oversight history for any AI system. Returns a signed chronological log for audit submissions. Filterable by date range, reviewer, and decision type.
// GET /v1/compliance/oversight/history?system_id=... // Optional: ?from=2026-01-01&to=2026-04-12&reviewer=human-rev-01 // Response { "total": 47, "events": [ { "log_id": "ovs-3a1f", "decision": "approved", "timestamp": "2026-04-12T14:00:00Z" } ], "audit_chain_valid": true }
Formally record serious incidents under EU AI Act
EU AI Act Article 73 requires providers of high-risk AI systems to notify authorities of serious incidents. Log and timestamp each incident with a signed formal record and severity classification.
// POST /v1/compliance/incident { "system_id": "model-prod", "severity": "serious", "description": "Biased output on protected group", "affected_users": 12 } // Response { "incident_id": "inc-8b4c", "timestamp": "2026-04-12T14:22:00Z", "notification_due": "2026-04-27T14:22:00Z", "theorem": "INC-100-IncidentBound" }
List and export your incident history
Query all recorded incidents for a system. Exports a signed incident registry suitable for regulatory submission. Filterable by severity, date, and resolution status.
// GET /v1/compliance/incidents?system_id=model-prod // Optional: ?severity=serious&from=2026-01-01 // Response { "total": 3, "incidents": [ { "incident_id": "inc-8b4c", "severity": "serious", "resolved": false } ], "registry_signed": true }
Auto-generated compliance transparency report
Generate a complete transparency report for any AI system, ready for board, regulator, or public disclosure. Covers decision distribution, fairness metrics, drift history, and oversight summary — all with signed theorem citations.
- ✓ Decision distribution and outcome metrics
- ✓ Fairness section with group-level breakdown
- ✓ Drift summary + DriftGuard certificate link
- ✓ PDF-ready + machine-readable JSON
- ✓ $0.08/report · 3 free · no API key needed
// Generate transparency report { "system_id": "loan-scorer-v3", "period": "2026-Q1", "include": ["fairness", "drift", "oversight"] } // Response: full signed report { "report_id": "trp-d1c7a9", "period": "2026-Q1", "total_decisions": 142830, "fairness_score": 0.89, "drift_certificates": 12, "oversight_events": 47, "theorem": "TRP-100-TransparencyBound", "pdf_url": "https://..." }
// Check for model drift with PSI bounds { "model_id": "model-prod-v2", "reference_window": "2026-01", "current_window": "2026-04" } // Response: PSI drift with formal bound { "drift_score": 0.08, "within_bound": true, "psi_bound": 0.20, "confidence": 0.95, "theorem": "DRG-100-DriftBound", "certificate_id": "drg-5a2b" } // GET /v1/drift/certificate?id=drg-5a2b // Returns: downloadable signed PDF certificate
Continuous drift detection with downloadable certificate
EU AI Act Article 72 mandates post-market monitoring. DriftGuard tracks Population Stability Index (PSI) across rolling windows and issues a signed certificate proving your model's behavior remains within the DRG-100-DriftBound theorem. Download the certificate to include in regulatory submissions.
- ✓ PSI drift score — formally bounded ≤ 0.20
- ✓ 95% confidence theorem: DRG-100-DriftBound
- ✓ Downloadable signed PDF certificate (/v1/drift/certificate)
- ✓ Real-time alerts on bound violation
- ✓ $0.01/check · 3 free · no API key needed
One API. Three frameworks.
All products map to EU AI Act, NIST AI RMF, and ISO 42001 simultaneously — no separate integrations.
Hand them a theorem.
Not a PDF.
Custom deployment, white-glove integration, dedicated rate limits, and co-authored compliance documentation available for enterprise customers. 24h response guaranteed.
atomadic@proton.me · Response within 24h · Custom SLAs available