AlgorithmWatch / Open Knowledge Foundation Germany OpenSCHUFA (2018)

In the spring of 2018 AlgorithmWatch and the Open Knowledge Foundation Germany launched OpenSCHUFA — the first widely-noticed European participatory-data-donation campaign on an algorithmic-accountability target. The campaign asked German residents to invoke their statutory subject-access right, free of charge, to obtain the credit-record data the dominant German credit-scoring bureau SCHUFA held on them, and to donate the resulting records to a public dataset that AlgorithmWatch and the Open Knowledge Foundation Germany would analyse in partnership with data journalists at the public broadcaster Bayerischer Rundfunk and Spiegel Online. Between May 2018 and 28 November 2018, when the analysis was published, the campaign generated more than 4,000 donated SCHUFA records, more than 30,000 subject-access requests to SCHUFA, and more than 100,000 subject-access requests across the German consumer-credit scoring field — by an order of magnitude the largest civic-side data-collection exercise on a single private-sector scoring system in European civil-society memory.

Origins and the Startnext crowdfunding

OpenSCHUFA was conceived inside AlgorithmWatch's first full year of operation — the organisation had been founded as a non-profit gGmbH in Berlin in 2017 — and was framed from the outset as the first concrete instantiation of AlgorithmWatch's wider working principle that empirical evidence about how specific automated decision-making systems actually behave is the precondition for democratic accountability over them. The choice of SCHUFA as the inaugural target was substantively heavy: SCHUFA's scores effectively gate rental tenancies, mobile-phone contracts, and consumer credit for some seventy million German adults; the company had operated with comparatively limited public scrutiny for decades; and the underlying scoring procedure was legally protected as a trade secret, leaving consumers with no statutory right to know how their score was calculated. To build the data-donation portal, AlgorithmWatch and the Open Knowledge Foundation Germany ran a Startnext crowdfunding round that raised more than €43,000 from more than 1,800 backers, a public mobilisation in its own right that was used by the campaign to argue that public concern about SCHUFA's operation was already widely diffused across the German consumer population and not confined to civil-society or academic researchers.

Methodology: statutory subject-access as the citizen-side lever

The campaign's methodological centre was the German statutory right to a free annual disclosure of personal data held by a data controller — under section 34 of the German Federal Data Protection Act (BDSG) at the campaign's launch, and from 25 May 2018 (three days after the public data-donation platform went live) under the directly applicable Article 15 of the European Union General Data Protection Regulation. The campaign actively encouraged participants to file subject-access requests not only with SCHUFA but also with Germany's other consumer-credit scoring firms — Boniversum, CRIF Bürgel, infoscore Consumer Data, and Deltavista — and to donate the SCHUFA records to the OpenSCHUFA public dataset for aggregate analysis. The openschufa.de portal supplied template letters, response-handling guidance, and a structured-upload interface that converted donated records into a single comparable dataset under AlgorithmWatch's and the Open Knowledge Foundation Germany's joint civil-society stewardship. AlgorithmWatch's responsible-party named on the campaign site's impressum was Matthias Spielkamp, AlgorithmWatch's co-founder and Executive Director. The data-journalism leg of the campaign was led by Bayerischer Rundfunk and Spiegel Online editors and data-journalists working alongside the campaign partners on the analysis of the donated records and on the public-facing reporting of the findings.

The November 2018 findings

On 28 November 2018 AlgorithmWatch and the Open Knowledge Foundation Germany — with parallel reporting by Spiegel Online and Bayerischer Rundfunk that same day — published the analysis of approximately 2,000 of the more than 4,000 donated SCHUFA records, the first systematic public examination of how a major European consumer-credit scoring system was operating in practice. The published findings substantiated two named anomalies in SCHUFA's scoring operation. First, the analysis identified cases in which "a number of people were rated rather negatively although SCHUFA had no negative information on them, e.g. on debt defaults" — that is, scores that diverged from the named risk-signals SCHUFA itself was claiming to use, suggesting that the operative scoring procedure was drawing on proxies and inferences whose presence in the model SCHUFA had never publicly disclosed. Second, the analysis identified divergences of up to ten per cent between SCHUFA's two simultaneously-operating scoring versions, with the older version 2 still in active commercial use alongside the newer version 3 even though SCHUFA's own internal assessment recognised the older version as inferior. The wider context — that the underlying scoring algorithm itself remained legally protected as a trade secret and was never disclosed by SCHUFA — meant that the campaign's findings were necessarily framed in terms of input-output anomalies and version-comparison artefacts rather than direct examination of the model. SCHUFA's separate decision to forbid the editors at Bayerischer Rundfunk and Spiegel Online to cite or present even a digest of SCHUFA's extensive written answers to the campaign partners' questions added a further documented opacity layer to the published reporting.

Public verdict and post-publication response

The findings prompted the German Consumer Affairs Council to demand that scoring agencies improve their information policy and provide more transparency, and prompted political demands inside the Bundestag and at federal-government level for legislative reform on credit-scoring transparency — the first concrete legislative-track pressure on the German credit-scoring sector since the BDSG's own scoring-transparency provisions had been adopted a decade earlier. The campaign's public verdict on its own outcome was mixed. The campaign had succeeded in establishing the citizens-as-data-donors architecture, in mobilising a participant population of substantial size (more than 4,000 donated records and more than 30,000 subject-access requests to SCHUFA), and in pressing the country's two most-read news outlets into running detailed empirical reporting on SCHUFA's operation. It had not, on the other hand, secured statutory disclosure of the scoring algorithm, nor moved the federal government to pass new primary legislation on credit-scoring transparency. AlgorithmWatch's openschufa.de post-mortem (last updated May 2019) framed the wind-down as a resource-constraint outcome: "Unfortunately, we lacked the resources to forcefully pursue the implementation of the demands we had stated", the campaign site reads.

Coalition shape and campaign repertoire

OpenSCHUFA's coalition architecture is structurally distinct from the other algorithmic-accountability campaigns in the corpus and is worth recording explicitly. The campaign was led jointly by two German civil-society organisations — AlgorithmWatch as the algorithmic-accountability advocacy-and-research lead, the Open Knowledge Foundation Germany as the open-data and civic-technology partner contributing platform-build and statutory-rights-mobilisation expertise — rather than by a single org with subordinate partners. Its operative coalition vehicle was the openschufa.de data-donation portal and the public dataset itself, with the Startnext crowdfunding round substituting for the conventional foundation-grant funding model of European civil-society campaigning. Its public-facing reach was channelled through two professional data-journalism newsrooms — Bayerischer Rundfunk as the German public-broadcaster anchor and Spiegel Online as the largest-circulation German online news outlet — rather than through a civil-society signature-block or coalition manifesto. And its substantive demand was structured as a public-data-and-anomalies-publication intervention rather than as a draft-law / petition / litigation campaign. The campaign repertoire that resulted — citizen-side statutory-rights mobilisation; crowdfunded platform build; aggregated donated-data analysis; data-journalism partnership for public reporting; post-publication political-pressure framing — is the working template that AlgorithmWatch has reused and adapted since: the 2020–2021 Instagram newsfeed monitoring shifted the participation interface from statutory-disclosure-and-donation to browser-add-on-and-scraping (and was shut down by Facebook on terms-of-service grounds in July 2021), and the 2020-ongoing DataSkop infrastructure reverted to a GDPR-mandated-data-export-and-donation architecture closer to OpenSCHUFA's, run as a BMBF-funded five-organisation consortium with election-year cycles on YouTube (2021 German federal election) and TikTok (2023).

Place in the make-AI-good movement

OpenSCHUFA matters to the wider make-AI-good corpus on three connected counts. First, it is the corpus's first sustained Continental European national-level algorithmic-accountability campaign and the corpus's first Germany-anchored campaign of any kind — closing a substantial geographic gap in the campaigns slice, which until this entry had Brussels-institutional EU campaigns (the EDRi-coordinated AI Act coalition), UK national campaigns, US class actions and federal-policy advocacy, Kenya-anchored content-moderation litigation, India biometric-surveillance (Project Panoptic), and a small set of international coalition campaigns, but zero Continental European national-level campaign anchored on a single national consumer-protection or financial-AI question. Second, the campaign sits in a movement area — consumer-credit and financial AI accountability — that the corpus has not previously had any campaign-side coverage of. The corpus's algorithmic-accountability cluster has anchored on UK welfare and immigration litigation (the Foxglove / JCWI visa-streaming, OFQUAL A-Level, and GMCDP / DWP work), US copyright class actions (Andersen v. Stability AI, the Authors Guild OpenAI class action), Kenyan content-moderation litigation (the Foxglove / Motaung Meta-Sama and 185 moderators cases), and EU AI-policy work — but had no campaign-side anchor on private-sector consumer-scoring systems whose operation directly shapes household financial outcomes for tens of millions of European consumers. Third, the participatory-data-donation methodology OpenSCHUFA established is the corpus's principal European working template for engaging non-AI publics directly in the empirical accountability work itself — distinct from the Algorithmic Justice League's Drag Your Feet and Freedom Flyers participatory-audit model, which is anchored on community-reporting of biometric-system encounters; distinct from the Internet Freedom Foundation's Project Panoptic crowdsourced "Report an FRT System" feature, which is a citizen-reporting extension on a desk-research-anchored public-tender database; and distinct from the Reclaim Your Face European Citizens' Initiative, which mobilises citizen signatures behind a coalition-led policy ask. OpenSCHUFA's working architecture — citizens invoke their statutory subject-access right, donate the resulting records to a public dataset, civil-society and journalist partners analyse the records, public anomalies-reporting feeds back into political and regulatory pressure on the controller — has been the corpus's most reused European participatory-research pattern across the seven years since.