What it is
Every record an organization keeps is a record it must protect, correct, and eventually delete. Data minimization asks what the work actually needs and keeps only that, for only as long as it is needed. The research cited below names the duty that comes with it. That duty is purpose limitation: consent a person gave for one purpose does not cover reusing their data to train a model for another. The same research names the cost, which is that a thinner record supports less checking.
What it pushes on in the Lab
In the Lab, this lever weakens five pathways. People and the automated system write less into records. People paste less client data into unsanctioned tools. The organization sends less identifiable data to a model that runs on an ungoverned host, and it copies fewer records outside the governed system.
You can also pull this lever at a strong tier, which costs more. At the strong tier, each effect below the Lab's strongest setting pushes harder.
In the modes that offer aiming, you can aim this lever at particular parts and pathways of a network. Otherwise it applies to the whole network.
This lever pushes less hard when you switch on lingering effects in the Lab's options, unless you also pull Understand the system.
Its pattern in the Practice Library
The Practice Library describes the pattern behind this lever:Data minimization
The pressures it answers
These pressures list this lever among the levers that answer them:
A lever answers a pressure when it pushes the other way on something the pressure pushes on.
Where you can pull it
Networks in the Lab that offer this lever:135
- A commercial code assistant across three enterprises
- Accelerated Safety Analysis Protocol (ASAP Tool)
- Advance Alert Monitor (AAM) deterioration model
- Albert France Services
- Allegheny Family Screening Tool
- Allegheny Hello Baby
- Allegheny Housing Assessment
- Amazon Flex driver standing and deactivation
Every network that offers it
- Amazon fulfillment-centre productivity discipline
- Amazon recruiting engine
- Ambient scribe RCT + monitoring playbook
- Amsterdam Smart Check
- An ambient AI scribe at a multi-specialty health system
- Aon's three-instrument pre-hire assessment suite
- Arkansas ARChoices / ARIA
- BAMF dialect recognition
- Benefits Data Trust wind-down
- BOSCO (Spain)
- Burokratt
- Caddy adviser copilot at Citizens Advice
- Calgary Drop-In Centre
- CDTFA Axyom Assist
- CHAI (chronic-homelessness prediction)
- Character.AI crisis-safety stack
- Checkr's automated background-check platform
- Cleveland State remote proctoring
- CNAF benefit-fraud risk score (France)
- Community Notes on X, formerly Birdwatch on Twitter
- CORA, the DC CFSA policy assistant
- CrimSAFE criminal-record tenant screening
- Crisis Text Line & Loris.ai
- Danske Bank fraud scoring
- Dave ExtraCash (CashAI)
- Douglas County Decision Aide
- DWP Whitemail Insights and Vulnerability Scanner
- Earnest AI underwriting
- EDD Virtual Assistant
- Enova's CashNetUSA and NetCredit loan servicing
- Epic Sepsis Model
- Family-Match (Adoption-Share)
- Fraud false positives that froze real accounts
- Frida (NAV Norway)
- Gaggle Safety Management
- GDS Microsoft 365 Copilot cross-government experiment
- GetCalFresh
- Gladsaxe model
- Google ML code completion
- Google's child-safety detection and account enforcement
- GOV.UK Chat
- Hackney / Xantura Early Help Profiling
- HireVue's video interview and assessment platform
- Home Office IPIC
- ID.me identity verification
- Illinois DCFS Augintel
- Illinois Rapid Safety Feedback
- Insight Bristol / Think Family Database
- INSS automated benefit analysis
- IRS collection chatbots
- iTutorGroup Tutor Application Screen
- Justice Transcribe
- Kaiser Permanente ambient AI scribe
- Kaiser Permanente Suicide-Risk Model
- LA County Homelessness Prevention Unit
- LA's coordinated-entry triage revision
- Learned Hand AI clerk pilot (LA and Riverside courts)
- Limbic Access (NHS Talking Therapies)
- London's Strategic Insights Tool
- Los Angeles County Project AURA
- LyssnCrisis counselor QA at ProtoCall Services (988)
- M-Shwari & Kenya's Digital Credit Market
- Magic Notes (Beam)
- Mass.gov Virtual Assistant
- Massachusetts DTA call summaries
- McHire, McDonald's franchise hiring platform
- Medicaid unwinding ex-parte renewals
- Meta employment-ad targeting and delivery optimization
- Meta's cross-check secondary review programme
- Michigan MiDAS
- Minute / Local Transcribe
- ML anti-money-laundering as primary monitoring
- MyFriendBen benefits screener
- NarxCare
- Nava assistive benefits chatbot
- Netherlands childcare-benefits scandal (Toeslagenaffaire)
- New Zealand MSD Predictive Risk Modelling
- nH Predict Utilization Review
- NJ AI Assistant
- NYC Teenspace
- ODMAP overdose spike alerts
- Oportun Financial Corporation's legal-collections pipeline
- Oregon Safety at Screening
- Oxevision camera monitoring on NHS mental health wards
- Predict-Align-Prevent
- ProKid (Netherlands)
- Propel in-app SNAP benefits assistant
- pymetrics Soft Skills Platform cooperative audit
- REACH VET
- RealPage revenue management
- Rotterdam welfare-fraud risk model
- Samagra Vedika
- San Jose's camera car
- Santa Clara County Homelessness Prevention System
- Sepsis Watch deep-learning detection system
- Serbia Social Card (Socijalna karta)
- Singapore's chatbot fleet refresh
- Sirius XM Radio's iCIMS-based applicant screening
- Sistema Alerta Niñez (Chile)
- SSA Insight
- StopNCII & Take It Down
- SyRI (Netherlands)
- Tessa chatbot replacing the NEDA eating-disorder helpline
- The Digit automated-savings tool, or Oportun Set & Save
- The GIFCT hash-sharing database and member matching system
- The NCMEC CyberTipline reporting and triage system
- The Sama Nairobi content-moderation workforce for Meta
- The same AI running hands off: the agentic office
- The same AI under full guardrails: the professional office
- The same AI with a human checking: the supervised office
- Trelleborg's Welfare Robot
- TREWS sepsis early-warning system
- Udbetaling Danmark data-driven control (Denmark)
- UK DWP Universal Credit Advances fraud model
- UK Home Office asylum AI copilots
- Unilever and HireVue graduate hiring
- UPS delivery route optimization
- Upstart lending model
- US Birth Match
- VA claims automation (automated survivor-benefit decisions)
- Vanderbilt VSAIL suicide-risk alert
- Viz.ai LVO Stroke Triage
- What Works for Children's Social Care ML pilots
- Wisconsin DEWS
- Woebot (a governed app wind-down)
- Workday AI screening
- Xantura OneView (predictive homelessness flagging)
The evidence behind its effects
The Lab cites these claims from the evidence registry for this lever's effects.
Minimisation carries a cost the protective case usually leaves out: a record deliberately kept thinner is also a record that supports less verification, so minimising trades exposure against the evidence the correction loop itself runs on. The duty that travels with it is purpose limitation — consent obtained for one purpose does not cover reuse of that data to train a model for another — which is how the data-protection regulation states the two together. Direction only: none of these sources measures the size of either cost.[3]
Downey, D. L., & Jenkins, D. A. (2026). AI in Supporting LGBTQIA+ Populations. In R. An & M. A. Lindsey (Eds.), Artificial Intelligence in Social Work: Bridging Technology and Humanity. Springer. https://doi.org/10.1007/978-3-032-18443-6_7
doi.org/10.1007/978-3-032-18443-6_7
Appears in: AI in Social Work (Springer, 2026)
Topics: lgbtqia, social-work
An, R., & Lindsey, M. A. (2026). Ethical Foundations of AI in Social Work. In R. An & M. A. Lindsey (Eds.), Artificial Intelligence in Social Work: Bridging Technology and Humanity. Springer. https://doi.org/10.1007/978-3-032-18443-6_2
doi.org/10.1007/978-3-032-18443-6_2
Appears in: AI in Social Work (Springer, 2026)
Topics: ai-ethics, social-work
European Parliament and Council of the European Union (2016). Regulation (EU) 2016/679 (General Data Protection Regulation), Article 5 — purpose limitation and data minimisation. https://eur-lex.europa.eu/eli/reg/2016/679/oj
https://eur-lex.europa.eu/eli/reg/2016/679/oj
Appears in: Evidence addition (2026)
Grounds: privacy law: purpose limitation and data minimisation (GDPR)
The LGBTQIA+ chapter documents a governance trade-off practitioners already make: social work professionals intentionally omit sexual-orientation and gender-identity data from client information systems to protect people from exposure, forced outing, or violence. The chapter frames this as a considered deviation from data-completeness norms rather than a recording error, reports it from the literature it reviews, and gives no prevalence figure.[†]
Downey, D. L., & Jenkins, D. A. (2026). AI in Supporting LGBTQIA+ Populations. In R. An & M. A. Lindsey (Eds.), Artificial Intelligence in Social Work: Bridging Technology and Humanity. Springer. https://doi.org/10.1007/978-3-032-18443-6_7
doi.org/10.1007/978-3-032-18443-6_7
Appears in: AI in Social Work (Springer, 2026)
Topics: lgbtqia, social-work
Tiered HIPAA penalties run from $145 to $73,011 per violation with an annual cap near $2.19M (2025-adjusted), and disclosure to a tool that is not a business associate is itself a violation.[2]
HIPAA Journal. (2026). HIPAA violation fines. https://www.hipaajournal.com/hipaa-violation-fines/
https://www.hipaajournal.com/hipaa-violation-fines/
Appears in: Paramerge authored research
Topics: privacy-security
HIPAA Journal. (2026). HIPAA violation penalties. HIPAA Journal.
Appears in: Paramerge authored research
Topics: privacy-security