10 to the 23 AI logo

Domain Atlas / Child welfare & family services

Case fileGladsaxe Municipality (Capital Region), Denmarkmedium deployment

Gladsaxe model

Work with this case in the PAN Lab ↗

Gladsaxe's early-detection project (DTO) was a decision-tree model over about 44 risk indicators, meant to score, for every child aged 0 to 6 rather than only families already receiving help, the estimated probability that the child was living in vulnerability; per a university-run Danish public-sector AI catalogue it was to be trained on roughly 173,000 notifications the authorities received between April 2016 and December 2017, but only about 117 usable historical cases existed, and it was halted in its development phase in 2019 without ever running on live decisions, after a national media storm and an unrelated data breach that exposed about 20,000 citizens' personal identification numbers.[5]

What happened

Gladsaxe, a Copenhagen-suburb municipality of about 70,000 people, built an in-house early-detection project (formally Dataunderstoettet Tidlig Opsporing af udsatte boern, or DTO) after a case review found children in distress were typically first identified only around age 8–12, even though different municipal departments had each noted earlier warning signs that were never connected into one picture. The design was a decision-tree model over roughly 44 risk indicators drawn from about nine cross-register data sources, meant to score — for every child aged 0–6 rather than only families already receiving help — the estimated probability that the child was living in vulnerability. Per a university-run Danish public-sector AI catalogue, only about 117 usable historical cases of vulnerable young children existed, too few to train on reliably, and the model was never completed. Contrary to durable public memory, the municipality and the peer-reviewed case study state it was not a "points system" that summed weighted points per risk factor; the widely quoted point values trace to media framing of the municipality's application. Human discretion was designed to remain decisive: a specialist adviser made the preliminary assessment and, without family consent, the case was to be deleted. The project sought legal cover first through a December 2017 "free municipality" exemption — denied by the Ministry of Economy and Interior — and then hoped for a basis in the new Data Protection Act, which did not, in the end, enable it. It became entangled in the national "ghetto" package and GDPR debates, and a March 2018 media storm compared it to surveillance states. After an unrelated December 2018 theft of four laptops exposed a spreadsheet with about 20,000 citizens' personal identification numbers, a governing coalition party withdrew support; the project was put on hold in January 2019 and ended in its development phase that year, never deployed on live decisions.

The sociotechnical reading

Most cases in this Atlas turn on what a deployed system did; Gladsaxe turns on what a system was never permitted to do. Its decisive controls all sat upstream of the model: a legal authorization it never obtained (a free-municipality exemption, then a data-protection legal basis, neither granted), a deployment sign-off, the whole-population scope of its data (it was to profile every young child, not only families already receiving help — a broader reach than the benefits-gated tools it was modeled on), and the institution's data-security capacity (a part-time data-protection officer against an estimated need of ten to seventeen). The model's own dynamics — the memory loop, the deference loop — barely got a chance to run. The case study's diagnosis is a "decoupled" governance chain: councillors approved the purpose but were not close enough to the method to explain or defend it when a media storm and an unrelated data breach collapsed the political support the project depended on. The lesson is that legitimacy, legal basis, and security capacity are load-bearing controls in their own right — a tool can be stopped by the failure of any one of them long before its accuracy is ever tested. It is also the Atlas's clearest "double ethics" case: the same reading that flags whole-population profiling has to weigh the documented cost of not connecting signals that left vulnerable children unseen until age eight.

The concepts used in this reading are defined in the Field Guide; the governance responses live in the Practice Library. The model organization for this case can be stress-tested in the PAN Lab.

Grounding sources for this case

The same sources that ground this model organization in the PAN library: evaluations, government documents, investigative reporting, and advocacy documentation, each labeled by tier.

kennethkristensensamfundsled2022GroundingAcademicSave

Kenneth Kristensen (Samfundslederskab i Skandinavien, Copenhagen Business School), Hvorfor Gladsaxemodellen fejlede: om anvendelse af algoritmer paa socialt udsatte boern (2022) https://rauli.cbs.dk/index.php/SiS/article/view/6542

https://rauli.cbs.dk/index.php/SiS/article/view/6542

Grounds: model org: gladsaxe_dto

offentligaiuniversityrundanindGroundingAcademicSave

Offentlig AI (university-run Danish public-sector AI catalogue), Gladsaxe-modellen project profile (n.d.) https://offentlig-ai.dk/projekter/gladsaxe-modellen

https://offentlig-ai.dk/projekter/gladsaxe-modellen

Grounds: model org: gladsaxe_dto

helenefriisratnerandkasperel2023GroundingAcademicSave

Helene Friis Ratner and Kasper Elmholdt, Algorithmic constructions of risk: Anticipating uncertain futures in child protection services, Big Data and Society (2023) https://journals.sagepub.com/doi/10.1177/20539517231186120

https://journals.sagepub.com/doi/10.1177/20539517231186120

Grounds: model org: gladsaxe_dto

Topics: algorithmic-fairness, child-welfare

katarinafastlappalainen2021GroundingAcademicSave

Katarina Fast Lappalainen, Protecting Children from Maltreatment with the Help of Artificial Intelligence: A Promise or a Threat to Children's Rights?, De Lege 2021 (Uppsala University Faculty of Law) (2021) https://www.diva-portal.org/smash/record.jsf?pid=diva2:1653453

https://www.diva-portal.org/smash/record.jsf?pid=diva2:1653453

Grounds: model org: gladsaxe_dto

Seeing your organization in this case file?

The histories here are documented after the harm. Mapping a live deployment's pathways and pressures, before the incident report, is engagement work: intake, diagnosis, prescription, and monitoring, with every limitation stated.

Sources & Evidence

Claims made on this page and what supports them. The full registry lives in Evidence.

EmpiricalGladsaxe's early-detection project (DTO) was a decision-tree model over about 44 risk indicators, meant to sco…

Gladsaxe's early-detection project (DTO) was a decision-tree model over about 44 risk indicators, meant to score, for every child aged 0 to 6 rather than only families already receiving help, the estimated probability that the child was living in vulnerability; per a university-run Danish public-sector AI catalogue it was to be trained on roughly 173,000 notifications the authorities received between April 2016 and December 2017, but only about 117 usable historical cases existed, and it was halted in its development phase in 2019 without ever running on live decisions, after a national media storm and an unrelated data breach that exposed about 20,000 citizens' personal identification numbers.

offentligaiuniversityrundanindGroundingAcademicSave

Offentlig AI (university-run Danish public-sector AI catalogue), Gladsaxe-modellen project profile (n.d.) https://offentlig-ai.dk/projekter/gladsaxe-modellen

https://offentlig-ai.dk/projekter/gladsaxe-modellen

Grounds: model org: gladsaxe_dto

kennethkristensensamfundsled2022GroundingAcademicSave

Kenneth Kristensen (Samfundslederskab i Skandinavien, Copenhagen Business School), Hvorfor Gladsaxemodellen fejlede: om anvendelse af algoritmer paa socialt udsatte boern (2022) https://rauli.cbs.dk/index.php/SiS/article/view/6542

https://rauli.cbs.dk/index.php/SiS/article/view/6542

Grounds: model org: gladsaxe_dto

helenefriisratnerandkasperel2023GroundingAcademicSave

Helene Friis Ratner and Kasper Elmholdt, Algorithmic constructions of risk: Anticipating uncertain futures in child protection services, Big Data and Society (2023) https://journals.sagepub.com/doi/10.1177/20539517231186120

https://journals.sagepub.com/doi/10.1177/20539517231186120

Grounds: model org: gladsaxe_dto

Topics: algorithmic-fairness, child-welfare

katarinafastlappalainen2021GroundingAcademicSave

Katarina Fast Lappalainen, Protecting Children from Maltreatment with the Help of Artificial Intelligence: A Promise or a Threat to Children's Rights?, De Lege 2021 (Uppsala University Faculty of Law) (2021) https://www.diva-portal.org/smash/record.jsf?pid=diva2:1653453

https://www.diva-portal.org/smash/record.jsf?pid=diva2:1653453

Grounds: model org: gladsaxe_dto