🇳🇱Gemeente Den Haag

Sociale Dienst Risk Scoring (Den Haag)

High RiskNetherlandsWelfareCommunityunverified deployment

A risk scoring system used by The Hague municipality social services to prioritise welfare claimants for additional verification checks based on machine learning predictions of non-compliance likelihood.

AI classification

Classified as high risk under EU AI Act Annex III. Governments must register systems like this publicly. Registration status: incomplete.

FOI requests

Responses received

Documents

Contributions

News mentions

FOI requests

Request re: algorithmic logic and data sources
Filed by Bits of Freedom · 14 Jan 2024
Responded
Request re: impact assessment documentation
Filed by Anonymous · 3 Mar 2024
Pending
Request re: vendor contract and procurement
Filed by Waag · 19 Nov 2023
Responded
Request re: automated decision appeal process
Filed by Clinic for Digital Rights · 2 Feb 2025
No response

AI analysis & document chat

AI-generated summary — FOI response · Bits of Freedom · Jan 2024

Based on uploaded document · 34 pages · always verify against original

AI summary

Disclosed

General system purpose, data categories used, list of municipalities where deployed.

Withheld

Algorithmic weighting logic, model accuracy metrics, training data sources.

Technical details

Risk score range 0–100. Threshold for manual review: 70+. Updated quarterly.

Refusal basis

Art. 10 Woo — third party commercial interests (vendor contract).

Ask about this document

You

Why was the algorithmic logic withheld?

The response cites Article 10 of the Wet open overheid (Woo), specifically the exemption protecting third-party commercial interests. The vendor's weighting methodology is considered a trade secret. This exemption can be challenged if you can demonstrate public interest outweighs commercial harm. The Dutch Administrative Court has precedent for overturning such refusals in algorithmic accountability cases.

Community, news & connections

Community contributions

System is still referenced in internal municipal guidelines despite the 2020 court ruling suspending deployment.

Bits of Freedom · verified · 12 Feb 2025 • verified

Vendor identified as Capgemini Netherlands based on procurement register cross-reference.

Waag · verified · 4 Dec 2023 • verified

I was flagged by this system in 2022 and denied housing benefit with no explanation given.

Anonymous citizen · unverified · 1 Mar 2024 • unverified

News mentions

NRC

Dutch welfare algorithm violated privacy rights, court rules auto-detected

5 Feb 2020

The Guardian

Netherlands bans fraud detection algorithm after rights violations

6 Feb 2020

Wired

SyRI ruling signals new era for algorithmic accountability in Europe auto-detected

14 Mar 2020

Politico

EU AI Act fallout: which government systems must now be registered?

22 Jan 2025

Related algorithm profiles

SCHUFA Score — Germany

Same domain · same vendor · 2 shared FOI findings

High risk

Fraud Risk Model — Estonia

Same domain · similar refusal pattern

High risk

ROBAS — Czech Republic

Same domain · Capgemini vendor match

Unclassified

Benefits Scoring Tool — Georgia

Same domain · community flagged

Unclassified

AI pattern detected

4 welfare scoring systems across the EU share the same vendor and a consistent pattern of withholding algorithmic logic under commercial interest exemptions. Refusal rate: 87%.

Accountability timeline

Feb 2025

FOI on appeal process — no response

Jan 2024

Algorithmic logic withheld — vendor exemption cited

Nov 2023

Procurement contract obtained via Waag FOI

Feb 2020

Court ruled SyRI violates privacy — deployment suspended