AI governance platform

From decisions to provable intelligence.

The governance and verification engine that turns every AI-driven decision into traceable, auditable, compliant evidence. Built for environments where mistakes are costly.

Decision record

DEC-2041 · Customer credit line

Verified
  1. 1Request receivedERP · credit department
  2. 2Rules applied4 of 4 passed · credit policy v3
  3. 3Model processingProduction model, unmodified
  4. 4Output verificationSources and thresholds checked
  5. 5Human reviewApproved by risk manager
Evidence recorded · sha256 · 7f9a…b84cAI Act Art. 12 · 14

Illustrative example

Non-invasive
No changes to your models or existing systems
Tamper-evident
Cryptographically sealed decision ledger
Standards-aligned
EU AI Act · ISO/IEC 42001 · NIST AI RMF
01The problem

Opaque AI doesn’t scale
in regulated markets.

AI models create value, but without an immutable record of how and why they reached a decision, their use in critical domains stays blocked by legal, regulatory and operational risk.

Decision accountability

When a model declines a credit line, flags a medical test or executes a transaction, your legal team must be able to show how and why. Without a verifiable trail, the decision cannot be defended.

Regulatory obligations and penalties

For high-risk systems, the EU AI Act requires continuous record-keeping, monitoring and human oversight (Articles 12 and 14). The most severe penalties reach €35 million or 7% of global annual turnover.

Unchecked drift

Models change with every update and parameter tweak. Without deterministic guardrails, small variations can cause material operational errors that go unnoticed.

02The solution

Verifiable
by design.

IRIS Core doesn’t modify your models. It wraps any AI model in a verification layer, making every decision it makes auditable and provable.

Decision traceability

Every decision leaves proof.

The input, context, instructions and output of every decision are recorded in real time with a cryptographic fingerprint, in an append-only ledger that cannot be altered.

  • A cryptographic fingerprint for every decision
  • Immutable, hash-chained ledger
  • Full reconstruction, even months later
Regulatory alignment

Regulations become controls.

Native mapping to the leading AI frameworks: EU AI Act (Articles 9, 12, 14, 15), ISO/IEC 42001, NIST AI RMF and industry standards.

  • Automated checks against AI Act articles
  • Continuous governance aligned with ISO/IEC 42001
  • Exportable compliance dossiers
Human oversight

People decide when it matters.

Programmable checkpoints let domain experts review, correct or approve critical outputs before they take effect.

  • Automatic escalation based on defined rules
  • Multi-party approval for high-impact decisions
  • Reviewer signature bound to the decision
03Modular architecture

Four modules.
One system.

Independent yet integrated modules: enable them as you need them, and they all work on the same decision ledger.

Module 01IRIS Trace

The decision ledger

Records every decision, instruction, context and output in real time in an immutable ledger, with millisecond precision.

10:45:12DEC-2041passed
10:45:11DEC-2040enforced
10:45:10DEC-2039logged
10:45:08DEC-2038passed

Illustrative interface

Module 02IRIS Policy Engine

Boundaries that hold

Programmable rules for compliance, ethical constraints and safety, enforced before and after every model call.

rule: human_oversight_art14

if risk > 0.72 or confidence < 0.85

then route to review · risk analysts

else approve and sign

Illustrative interface

Module 03IRIS Evidence Generator

Audit-ready dossiers

Compiles technical, contextual and operational evidence into dossiers readable by people and machines, for internal audit and regulators.

AI Act dossier · Art. 14PDF · JSON-LD
Classification
High risk
Human oversight
Active
Seal
0x4a9e…

Illustrative interface

Module 04IRIS Discovery

Retrospective analysis

Visual exploration of past decisions: anomalies, alternative scenarios and reconstruction of reasoning paths.

Illustrative interface

04Integration

A governance layer
you never see.

IRIS Core sits along the path of every decision: it checks before, verifies after and records every step. Your systems and models stay exactly as they are.

  1. 01Application

    Request

    The request originates in your systems (API, ERP, service).

  2. 02IRIS

    Rule check

    Company policies and regulations are checked before the model is called; context is recorded.

  3. 03Model

    Processing

    Any model: proprietary, on-premises or cloud. IRIS never modifies it.

  4. 04IRIS

    Output verification

    Consistency, sources and thresholds are checked; signed evidence goes to the ledger.

  5. 05Application

    Verified response

    The decision arrives with its proof, ready for audit and review.

Any model

Proprietary, on-premises or cloud models.

Legacy systems too

Runs alongside ERP and AS/400 with no automated writes.

Immutable ledger

Every step produces verifiable evidence.

05Industries

For environments
where mistakes are costly.

Industries where compliance is non-negotiable, errors carry legal consequences and black-box automation is simply not acceptable.

Financial services

Credit, fraud prevention, automated operations

Credit assessments, anti-money-laundering checks and operational decisions with explanations that hold up under regulatory inspection.

Reference use case

Explaining a declined credit application with documented alternatives.

SR 11-7 · Basel III

Healthcare and life sciences

Clinical protocols and diagnostic support

Diagnostic support and triage, with a clear, always-visible link between clinical guidelines and each recommendation.

Reference use case

Linking every recommendation to its underlying scientific sources.

EU MDR · FDA AI/ML guidance

Public sector and legal

Administrative decisions and regulatory summaries

Administrative determinations and regulatory summaries with precise citations and clear explanations for citizens.

Reference use case

Legal summaries with citations verified down to article and paragraph.

AI Act · high-risk categories

Enterprise governance

Visibility across every model in use

A single view for the board, general counsel and risk committees: models, vendors, drift and policy violations.

Reference use case

A compliance dashboard covering drift, costs and violations.

ISO/IEC 42001 · NIST AI RMF