Demo dashboard — sample data, not a live tenant.

Interactive demo

Northwind Financial (a demo organisation)

A mid-sized retail bank invented for this demo. It runs five AI systems across lending, hiring, payments and service. Assessed against the EU AI Act (Regulation 2024/1689), corpus 2026.08.28.

Northwind Financial is a fictional organisation and its five AI systems are invented. No customer data appears here. The risk classifications and obligations below are computed live by the same deterministic, citation-pinned engines the paid platform runs, from the sample systems described. They are an illustration for qualified professional review, not a legal determination.

AI systems in scope

Registered in this sample estate

High risk

Classified by the risk engine, not declared

Evidence complete

3 of 5

Every artefact the tier requires is locked

Open actions

Required artefacts not yet produced

Compliance score

80%

Tier-weighted, locked evidence only

How a system moves through the platform

Filing route shown for the EU AI Act (Regulation 2024/1689)

Classify

Each registered system is classified from what it does.

Module 1.1 risk classification and the Module 1.1b prohibition screen.

Prove

The obligations that follow are turned into citation-pinned artefacts.

Impact assessment, transparency notice and the evidence register, each content-hashed.

Monitor

Live systems are watched for drift and posture change.

Module 3.1 drift monitoring, with Sentinel on the Enterprise and Sovereign tiers.

File

A reportable incident is routed to the national market surveillance authority of the affected Member State.

EU AI Act Art. 73 (serious-incident reporting). Ongoing monitoring sits under EU AI Act Art. 72 (post-market monitoring).

React Flow

AI system inventory

Risk class computed by Module 1.1 from each system's description

System	Owner	Risk class	Modules run	Evidence
Candidate screening assistant NWF-HR-01 Ranks and shortlists job applicants for interview from their CV and application form. Used by the recruitment team for every open role in the group.	Group HR	High risk Purpose text implies Annex III 4 employment/worker management	1.1 · 1.1b · 1.2 · 1.3	Evidence complete 3 locked artefacts
Consumer credit decision engine NWF-CR-02 Produces a credit score for retail loan applications and sets the credit limit offered to an approved applicant. Runs on every personal loan and card application.	Retail lending	High risk Purpose text implies Annex III 5(b) creditworthiness	1.1 · 1.1b · 1.2 · 1.3 · 2.1	1 outstanding 2 locked artefacts; outstanding: transparency notice
Payment fraud detection NWF-FR-03 Screens card and transfer activity as it happens and flags suspected fraud for a human investigator to review before any account action is taken.	Financial crime	Limited risk systematic monitoring without significant effects	1.1 · 1.1b · 3.1	1 outstanding 1 locked artefact; outstanding: transparency notice
Customer support assistant NWF-CS-04 Answers customer questions about products, balances and branch services in a chat window, and hands the conversation to an adviser when it cannot answer.	Service operations	Minimal risk	1.1 · 1.1b · 1.3	Evidence complete 2 locked artefacts
Branch demand forecasting NWF-OP-05 Forecasts counter and call-centre volume by branch and hour so rotas can be planned a fortnight ahead. Reads aggregated volume counts only.	Operations planning	Minimal risk	1.1 · 1.1b	Evidence complete 1 locked artefact

Compliance posture

Scored with the same tier-aware rule the platform applies inside a tenant: one point per artefact the system's risk class requires, earned only when a content-hashed, current document satisfies it.

80%

across 5 systems, 45 obligations mapped

Systems classified5 of 5

Evidence complete for tier3 of 5

Transparency duty engaged1 of 5

Obligations and filing status

Derived from the EU AI Act (Regulation 2024/1689) by the policy engine

Each control below is pinned to a corpus citation. The platform derives them from the computed risk class, so a re-classification changes the obligation set rather than leaving a stale checklist behind.

Candidate screening assistantHigh riskEvidence complete20 controls mapped
Full provider obligations apply before it can be relied on.
- eu-ai-act-art-14 Implement a human-oversight control that can override or reverse this system's output.
- eu-ai-act-art-6 Document the Annex III mapping and any Article 6(3) 'no significant risk' exemption rationale for this system.
- eu-ai-act-art-9 Maintain a documented, lifecycle-long risk-management system for this high-risk system (Art 9).
- eu-ai-act-art-10 Evidence data-governance and bias controls over this system's datasets (Art 10).
- eu-ai-act-art-11 Assemble and maintain the Annex IV technical-documentation dossier for this system (Art 11).
and 15 further mapped controls, each pinned to its own citation.
Consumer credit decision engineHigh risk1 outstanding20 controls mapped
Full provider obligations apply before it can be relied on.
- eu-ai-act-art-14 Implement a human-oversight control that can override or reverse this system's output.
- eu-ai-act-art-6 Document the Annex III mapping and any Article 6(3) 'no significant risk' exemption rationale for this system.
- eu-ai-act-art-9 Maintain a documented, lifecycle-long risk-management system for this high-risk system (Art 9).
- eu-ai-act-art-10 Evidence data-governance and bias controls over this system's datasets (Art 10).
- eu-ai-act-art-11 Assemble and maintain the Annex IV technical-documentation dossier for this system (Art 11).
and 15 further mapped controls, each pinned to its own citation.
Payment fraud detectionLimited risk1 outstanding1 control mapped
Transparency obligations apply.
- eu-ai-act-art-4 Run and evidence an AI-literacy programme for staff and operators handling this system, proportionate to their role and the deployment context (Art 4).
Customer support assistantMinimal riskEvidence complete3 controls mapped
No mandatory high-risk obligations triggered. Good practice still applies.
- eu-ai-act-art-50 Disclose AI interaction to end users; mark generated content as artificially produced.
- eu-ai-act-art-53 Maintain GPAI technical documentation and training-data summary; add systemic-risk controls where Article 51 thresholds are met (Art 53–55).
- eu-ai-act-art-4 Run and evidence an AI-literacy programme for staff and operators handling this system, proportionate to their role and the deployment context (Art 4).
Branch demand forecastingMinimal riskEvidence complete1 control mapped
No mandatory high-risk obligations triggered. Good practice still applies.
- eu-ai-act-art-4 Run and evidence an AI-literacy programme for staff and operators handling this system, proportionate to their role and the deployment context (Art 4).

Sentinel monitoring

Enterprise and Sovereign, closed beta

Monitored

On drift watch

Open incidents

Sample signals for a demo organisation. Sentinel is a closed beta and is not open for enrolment, so no estate is monitored today and these counts are invented. Ongoing monitoring sits under EU AI Act Art. 72 (post-market monitoring). A reportable incident is designed to route to the national market surveillance authority of the affected Member State under EU AI Act Art. 73 (serious-incident reporting), no later than 15 days after awareness (EU AI Act Art. 73(2)).

How Sentinel works

Now run one of your own

The sandbox puts your own AI system through the same engines that produced everything above: the risk classification, the prohibited-practice screen, the obligations that follow and a sample of the evidence. No account, and what you type is not stored.

Assess your own AI system See a plan