[{"data":1,"prerenderedAt":666},["ShallowReactive",2],{"uc-benefit-fraud-and-error-detection":3,"uc-regulations":457},{"useCase":4,"evidence":209,"blitsAiDeployments":343,"benchmarks":344,"indicative":351,"related":354,"indexability":455,"includeUnpublished":215},{"title":5,"shortTitle":6,"seoTitle":7,"metaDescription":8,"status":9,"definition":10,"aliases":11,"industries":16,"functions":18,"patterns":22,"channels":25,"audience":28,"autonomy":29,"adoptionStage":30,"problem":31,"problemStats":32,"howItWorks":38,"valueDrivers":39,"kpis":43,"indicativeValue":48,"macroEstimates":77,"feasibility":78,"implementation":90,"risk":133,"blitsAi":186,"faq":188,"related":198,"datePublished":204,"dateModified":204,"lastVerified":204,"changelog":205,"slug":208},"AI for benefit fraud and error detection in social security","Benefit fraud and error detection","AI for benefit fraud and error detection","Risk models help benefits agencies pick whom to check. DWP says its model is 2.5 times more effective than random checks; Rotterdam stopped its model in 2022.","published","Risk models that help a social security or benefits agency decide which claims, payments and recipients to check for fraud or error, so that caseworkers verify the riskiest cases first, while every decision on entitlement stays with a person and the model is tested for fairness before and during use.",[12,13,14,15],"welfare fraud detection","benefit fraud risk scoring","social security fraud and error analytics","improper payments detection",[17],"government",[19,20,21],"fraud-prevention","citizen-services","case-management",[23,24],"prediction-and-scoring","anomaly-detection",[26,27],"api","internal-tools","back-office","assist","early-adopters","Benefits agencies pay very large sums to millions of people, and some of it goes to the wrong\nplace: organised fraud, individual misrepresentation, and honest mistakes by claimants or by the\nagency itself. Checking everyone is impossible and would delay payments to people who need them\nurgently, so agencies have to choose whom to check.\n\nThat choice is where automated risk scoring has done serious harm. The Dutch childcare benefits\nscandal, the SyRI welfare fraud system that a Dutch court stopped in 2020, Rotterdam's welfare risk\nmodel and Australia's Robodebt scheme each show one or more of the same failures: selection on\nnationality or on proxies for it, systems too opaque to check or challenge, and a burden of proof\nshifted onto people who were then treated as if they owed money or had committed fraud. Under the EU AI Act, systems that evaluate eligibility for public\nassistance, or grant, reduce, revoke or reclaim it, are high risk. The job is not only to catch\nfraud but to do so lawfully, proportionately and transparently.",[33],{"statement":34,"sourceTitle":35,"sourceUrl":36,"year":37},"The UK Department for Work and Pensions paid 1.3 million Universal Credit advances worth GBP 700 million in 2025 to 2026 and estimates the fraud and error in them at between GBP 20 million and GBP 90 million.","Effectiveness Assessment of Universal Credit Advances Model","https://www.gov.uk/government/publications/effectiveness-assessment-including-statistical-analysis-of-the-universal-credit-advances-machine-learning-model-1-april-2025-to-31-march-2026/effectiveness-assessment-of-universal-credit-advances-model",2026,"1. **Define the risk precisely.** A model targets one defined risk at one point in the process,\n   for example an advance request before payment or a specific change in circumstances, not\n   \"fraud\" in general.\n2. **Score at the point of decision.** The claim data is scored in real time or in batch. The\n   output is a referral for a check, never a decision on entitlement.\n3. **Blind and mix the referrals.** High risk referrals go to caseworkers together with a random\n   control group, and the caseworker is not told which is which, so the check is not biased by the\n   score.\n4. **Verify with the person.** A caseworker reviews the evidence and, where needed, asks the\n   claimant, then decides. Declines follow the normal notice, review and appeal route.\n5. **Measure effectiveness and fairness.** Confirmed fraud and error rates in model referrals are\n   compared with the random group, overall and by group, and the model is retrained or stopped when\n   it targets groups without a matching rate of confirmed findings.",[40,41,42],"risk-reduction","compliance","cost-to-serve",[44,45,46,47],"detection-rate-improvement","false-positive-reduction","fraud-loss-reduction","cost-savings",{"referenceOrg":49,"inputs":50,"formula":72,"currency":73,"period":74,"resultLabel":75,"caveat":76},"A national benefits agency paying EUR 2 billion a year in a benefit with known fraud and error risk",[51,58,65],{"key":52,"label":53,"low":54,"high":55,"unit":56,"note":57},"paymentsValue","Value of payments in scope",1000000000,3000000000,"EUR per year","Editorial assumption. Replace with the payments the model will actually screen.",{"key":59,"label":60,"low":61,"high":62,"unit":63,"note":64},"lossRate","Share of payments lost to fraud and error",0.01,0.03,"fraction of payments","Editorial assumption. Use your own official fraud and error statistics, which vary widely by benefit.",{"key":66,"label":67,"low":68,"high":69,"unit":70,"note":71},"preventedShare","Share of those losses prevented by better targeted checks",0.05,0.15,"fraction of losses","Editorial assumption, deliberately modest. DWP reports its advances model is 2.5 times more effective than random selection, but a better hit rate prevents only part of the loss.","paymentsValue * lossRate * preventedShare","EUR","per year","Fraud and error losses prevented","Gross losses prevented only. It leaves out the cost of the checks, the cost of wrongly delayed or refused payments to legitimate claimants, legal and reputational risk, and the cost of the governance a high risk system requires.",[],{"complexity":79,"complexityNote":80,"dataPrerequisites":81,"integrations":85},"high","Technically this is a scoring model; in practice it is one of the most regulated and contested uses of AI in government. It needs a clear legal basis, a data protection impact assessment, an equality or fundamental rights assessment, published documentation and an evaluation design with a random control group.",[82,83,84],"Confirmed outcomes of past checks, including checks that found nothing","Claim data available at the decision point, with a legal basis for each item","Protected characteristic data or reliable proxies for fairness testing",[86,87,88,89],"Benefit claim and payment systems","Caseworker case management","Notice, review and appeal processes","Data warehouse for monitoring and evaluation",{"steps":91,"guardrails":107,"humanInTheLoop":113,"kpisToInstrument":114,"failureModes":120},[92,95,98,101,104],{"title":93,"detail":94},"Choose a narrow, well evidenced risk","Start with one payment type where fraud is documented, as DWP did with Universal Credit advances, rather than a general score of every recipient.",{"title":96,"detail":97},"Do the rights assessment first","Complete the data protection impact assessment and an equality or fundamental rights impact assessment before build, and publish a transparency record.",{"title":99,"detail":100},"Build the evaluation into the design","Keep a random control group of referrals and hide the source of each referral from caseworkers, so effectiveness and fairness can be measured honestly.",{"title":102,"detail":103},"Test disparities in referral and in outcome","For every group you can measure, compare how often the model refers people with how often those referrals are confirmed. DWP's assessment found groups referred more often without more confirmed fraud, and DWP retrained the model, which was still in testing when it reported.",{"title":105,"detail":106},"Define the stop rule","Agree in advance what finding would pause the model. Rotterdam stopped its welfare model when it concluded it could not currently build a model that fits its policy.",[108,109,110,111,112],"The model only refers cases for a check; it never decides, reduces or stops a payment","No protected characteristic, nationality or obvious proxy as a feature","Random control group and blinded referrals for every model in use","Published transparency record and effectiveness assessment at least yearly","Payment timeliness for legitimate claimants monitored as a harm measure","Caseworkers review every referral and make every entitlement decision, with the claimant able to explain their circumstances. Claimants keep the normal review and appeal rights. A senior owner signs off the yearly effectiveness and fairness assessment and can suspend the model.",[115,116,117,118,119],"Confirmed fraud and error rate in model referrals versus the random control group","Referral and outcome disparities by group","Payment delay for legitimate claimants who were referred","Share of declines overturned on review or appeal","Losses prevented per check",[121,124,127,130],{"title":122,"detail":123},"Discrimination through proxies","Nationality is used directly, or language, neighbourhood or family status stand in for protected characteristics. The Dutch childcare benefits model used nationality as a risk factor, and journalists found Rotterdam's model scored on age, gender and language skills.",{"title":125,"detail":126},"Automation reversing the burden of proof","When a score or data match is treated as proof, people must disprove a debt. Robodebt raised debts from averaged income data without other evidence and put the onus on recipients to contradict them. A person must establish the facts.",{"title":128,"detail":129},"Secrecy that blocks accountability","Agencies that refuse to disclose how their models work cannot be challenged or corrected, as investigations in Sweden and Denmark have shown. Publish the documentation.",{"title":131,"detail":132},"Chasing small sums at high human cost","Aggressive thresholds generate many checks on honest claimants for little recovered money. Measure the harm side alongside the savings.",{"euAiAct":134,"regulations":136,"guidance":143,"controls":155,"incidents":161},{"tier":79,"basis":135},"Annex III point 5(a): AI systems used by or on behalf of public authorities to evaluate the eligibility of natural persons for essential public assistance benefits and services, or to grant, reduce, revoke or reclaim them. A fundamental rights impact assessment (Article 27) is required before a public body deploys it. A design that scores people over time on their social behaviour or personal characteristics and leads to unrelated or disproportionate detrimental treatment would fall under the Article 5(1)(c) prohibition on social scoring.",[137,138,139,140,141,142],"eu-ai-act","gdpr","nist-ai-rmf","iso-42001","uk-gdpr","uk-atrs",[144,150],{"title":145,"issuer":146,"region":147,"url":148,"note":149},"Algorithmic Transparency Recording Standard hub","UK government","europe","https://www.gov.uk/government/collections/algorithmic-transparency-recording-standard-hub","UK central government bodies publish records of their algorithmic tools here, including DWP's record for its Universal Credit advances model.",{"title":151,"issuer":152,"region":147,"url":153,"note":154},"Algoritmeregister van de Nederlandse overheid","Government of the Netherlands","https://algoritmes.overheid.nl/nl","Dutch government bodies register algorithms used in benefit control here, for example UWV's unemployment benefit risk scan and Rotterdam's welfare risk model, which is listed as no longer in use.",[156,157,158,159,160],"Fundamental rights and data protection impact assessments before deployment","Yearly published effectiveness and fairness assessment with a random control group","Blinded referrals and human decision on every case","Documented stop rule and senior owner with authority to suspend","Notice, review and appeal routes that do not depend on knowing the model exists",[162,166,170,174,178,182],{"title":163,"url":164,"note":165},"Amnesty International: Xenophobic machines, the Dutch childcare benefits scandal","https://www.amnesty.org/en/documents/eur35/4686/2021/en/","The Dutch tax authorities used an algorithmic system to create risk profiles of childcare benefit applicants, with nationality as one of its risk factors. Amnesty found this resulted in discrimination and racial profiling.",{"title":167,"url":168,"note":169},"District Court of The Hague: SyRI judgment (ECLI:NL:RBDHA:2020:865)","https://uitspraken.rechtspraak.nl/inziendocument?id=ECLI:NL:RBDHA:2020:865","The court ruled in February 2020 that the legislation for SyRI, a Dutch government system that linked data to flag welfare fraud risk, breached the right to private life because it was insufficiently transparent and verifiable.",{"title":171,"url":172,"note":173},"Royal Commission into the Robodebt Scheme: report volume 1 (archived)","https://web.archive.org/web/20241231202325/https://robodebt.royalcommission.gov.au/system/files/2023-07/robodebt_report_volume_1.pdf","Australia's Robodebt scheme raised welfare debts largely through income averaging, without other evidence, and placed the onus on recipients to contradict the result. The Royal Commission found the method neither produced accurate results nor complied with the income calculation provisions of the Social Security Act 1991. In 2020 the government decided to refund debts raised wholly or partly through averaging that had been repaid, and to reduce unpaid ones to zero, and in November 2020 it settled a class action. The report was presented on 7 July 2023. It was automation rather than machine learning, but the lesson on burden of proof applies.",{"title":175,"url":176,"note":177},"Lighthouse Reports: Suspicion Machines (Rotterdam)","https://www.lighthousereports.com/investigation/suspicion-machines/","Reconstruction of Rotterdam's welfare fraud model, which took 315 inputs including age, gender and language skills; the investigation found it discriminated by ethnicity, age, gender and parenthood.",{"title":179,"url":180,"note":181},"Lighthouse Reports: Sweden's Suspicion Machine","https://www.lighthousereports.com/investigation/swedens-suspicion-machine/","Analysis of the Swedish Social Insurance Agency's fraud prediction algorithm for temporary child support found it disproportionately flagged women, migrants, low income earners and people without a university education.",{"title":183,"url":184,"note":185},"Amnesty International: Denmark's AI powered welfare system fuels mass surveillance","https://www.amnesty.org/en/latest/news/2024/11/denmark-ai-powered-welfare-system-fuels-mass-surveillance-and-risks-discriminating-against-marginalized-groups-report/","Amnesty's 2024 report on the fraud control algorithms of Denmark's welfare authority, Udbetaling Danmark, warns of mass surveillance and discrimination against marginalised groups, including through inputs on \"foreign affiliation\". The authority refused full access to the code and data.",{"howToBuild":187},"Blits.ai is not a fraud scoring engine and should not be used to decide entitlement. Where it\nhelps is around the model: an **agentic workflow** can take a referral, gather the claim facts\nthrough **custom functions**, prepare a neutral case summary for the caseworker and draft the\ninformation request to the claimant, with **human in the loop approval** before anything is\nsent. A **knowledge base** with hybrid retrieval over the benefit rules and the evidence policy\nhelps caseworkers apply the rules consistently.\n\n**PII masking** at the gateway, a full audit trail per run and **test suites** that check that\nsummaries stay neutral and factual support the documentation a high risk use needs. The platform\nis **model agnostic** and offers EU and UAE data residency for claimant data.",[189,192,195],{"question":190,"answer":191},"Is benefit fraud detection high risk under the EU AI Act?","Yes, when the system evaluates eligibility for public assistance or is used to grant, reduce, revoke or reclaim benefits (Annex III point 5(a)). Public bodies must also carry out a fundamental rights impact assessment before use.",{"question":193,"answer":194},"Does it work?","It can improve targeting. DWP reports its Universal Credit advances model is 2.5 times more effective than random selection in 2025 to 2026, with a median payment delay of one day for approved referrals. The same assessment found disparities by nationality, age and couple status, which is why the random control group and yearly assessment matter.",{"question":196,"answer":197},"What went wrong in the Dutch, Australian and Rotterdam cases?","Selection used nationality or proxies such as language skills, people had to disprove a score or data match, and systems were hard to scrutinise. Rotterdam stopped its model in 2022, a Dutch court struck down the SyRI legislation in 2020, and debts Robodebt raised through income averaging were refunded or reduced to zero before a Royal Commission reported in 2023.",[199,200,201,202,203],"tax-compliance-risk-scoring","benefits-eligibility-and-application-assistant","inspection-prioritization","application-and-identity-fraud-detection","ai-model-inventory","2026-09-27",[206],{"date":204,"note":207},"First published","benefit-fraud-and-error-detection",[210,249,278,304,322],{"title":211,"useCases":212,"organization":213,"vendors":217,"summary":221,"stage":222,"year":37,"channels":223,"languages":224,"metrics":226,"outcomeDisclosed":235,"sources":236,"verification":243,"grade":246,"id":247,"organizationSlug":248},"Department for Work and Pensions: machine learning risk model for Universal Credit advances",[208,202],{"name":214,"anonymized":215,"country":216,"region":147,"industry":17},"Department for Work and Pensions",false,"GB",[218],{"name":219,"role":220},"In house (Integrated Risk and Intelligence Service)","in-house","DWP scores requests for Universal Credit advances in real time with a supervised machine learning classifier and refers the highest risk requests to a caseworker before payment. The caseworker is not told the referral came from the model, a random control group is referred alongside, and every decision to decline is made by a person and can be appealed. DWP's published effectiveness assessment for April 2025 to March 2026 finds the model 2.5 times more effective than random selection with a median payment delay of one day for approved referrals. It also finds that non UK nationals and several age bands were referred more often without a matching increase in confirmed fraud, and that referrals of couples were less often confirmed than those of single claimants; a retrained model was being tested.","scaled",[26],[225],"en",[227],{"kpi":44,"value":228,"unit":229,"qualifier":230,"period":231,"baseline":232,"claimant":233,"quote":234,"sourceUrl":36},2.5,"multiplier","exact","1 April 2025 to 31 March 2026","Randomised control group sample of advances","organization","The performance information for 2025 to 2026 demonstrates the model is 2.5 times more effective at identifying fraud risk than a randomised control group sample.",true,[237,239],{"url":36,"title":35,"publisher":214,"date":238},"2026-07-09",{"url":240,"title":241,"publisher":242},"https://www.gov.uk/algorithmic-transparency-records/dwp-universal-credit-advances-model","DWP: Universal Credit Advances Model (algorithmic transparency record)","GOV.UK",{"level":244,"checkedAt":245},"source-verified","2026-09-26","B","dwp-universal-credit-advances-fraud-model",null,{"title":250,"useCases":251,"organization":252,"vendors":256,"summary":259,"stage":222,"year":260,"channels":261,"languages":262,"metrics":263,"outcomeDisclosed":235,"sources":271,"verification":276,"grade":246,"id":277,"organizationSlug":248},"US Treasury: machine learning in federal payment fraud prevention and check fraud recovery",[208],{"name":253,"anonymized":215,"country":254,"region":255,"industry":17},"U.S. Department of the Treasury, Bureau of the Fiscal Service","US","north-america",[257],{"name":258,"role":220},"In house (Office of Payment Integrity)","The Office of Payment Integrity in Treasury's Bureau of the Fiscal Service added machine learning and risk based screening to how it checks federal payments. Treasury reports that these enhanced processes prevented and recovered over USD 4 billion in fraud and improper payments in fiscal year 2024, up from USD 652.7 million the year before. The release names machine learning only for identifying Treasury check fraud, which led to USD 1 billion in recovery; the USD 2.5 billion it reports from identifying and prioritising high risk transactions is not described as machine learning.",2024,[],[225],[264],{"kpi":265,"value":54,"unit":266,"currency":267,"qualifier":230,"period":268,"claimant":233,"quote":269,"sourceUrl":270},"fraud-losses-prevented","currency","USD","fiscal year 2024 (October 2023 to September 2024)","Expediting the identification of Treasury check fraud with machine learning AI resulting in $1 billion in recovery.","https://home.treasury.gov/news/press-releases/jy2650",[272],{"url":270,"title":273,"publisher":274,"date":275},"Treasury Announces Enhanced Fraud Detection Processes, Including Machine Learning AI, Prevented and Recovered Over $4 Billion in Fiscal Year 2024","U.S. Department of the Treasury","2024-10-17",{"level":244,"checkedAt":245},"us-treasury-payment-integrity-machine-learning",{"title":279,"useCases":280,"organization":281,"vendors":284,"summary":288,"stage":289,"year":290,"channels":291,"languages":292,"metrics":294,"outcomeDisclosed":215,"sources":295,"verification":302,"grade":246,"id":303,"organizationSlug":248},"Gemeente Rotterdam: welfare reassessment risk model, stopped in 2022",[208],{"name":282,"anonymized":215,"country":283,"region":147,"industry":17},"Gemeente Rotterdam","NL",[285],{"name":286,"role":287},"Accenture","integrator","From 2017 the City of Rotterdam used a machine learning model that gave each social assistance recipient a risk score between 0 and 1 for receiving benefits they were not, or no longer, entitled to, based on the outcomes of earlier eligibility reviews. High scores were one route to an invitation for a review interview, and an income consultant decided the outcome. The city classified the model as high risk and stopped using it in early 2022; its register entry states that a review found it was not currently possible to build a risk model that fits the city's policy. The register says the model processed no nationality, age or health data; journalists who obtained the model file reported 315 inputs, including age, gender and language skills, and found that it discriminated by ethnicity, age, gender and parenthood.","paused",2022,[27],[293],"nl",[],[296,299],{"url":297,"title":298,"publisher":151},"https://algoritmes.overheid.nl/nl/algoritme/gm0599/36585638/heronderzoeken-uitkeringsgerechtigden","Heronderzoeken Uitkeringsgerechtigden, Algoritmeregister",{"url":176,"title":300,"publisher":301},"Suspicion Machines","Lighthouse Reports",{"level":244,"checkedAt":245},"gemeente-rotterdam-welfare-reassessment-risk-model",{"title":305,"useCases":306,"organization":307,"vendors":309,"summary":310,"stage":311,"year":290,"channels":312,"languages":313,"metrics":314,"outcomeDisclosed":215,"sources":315,"verification":320,"grade":246,"id":321,"organizationSlug":248},"UWV: risk scan for culpable unemployment in unemployment benefit claims",[208],{"name":308,"anonymized":215,"country":283,"region":147,"industry":17},"Uitvoeringsinstituut Werknemersverzekeringen (UWV)",[],"The Dutch employee insurance agency UWV uses a risk scan on applications for unemployment benefit (WW) to signal cases where the applicant may have become unemployed through their own fault, which would remove the entitlement. The scan combines several data points, never a single characteristic, and uses no personal characteristics such as origin, gender or age; high risk applications are offered to staff for a fuller investigation, and staff decide. Specialists check the input data quality monthly and whether the development population is still representative, and an independent party reviews the work when the scan is developed further. Randomly chosen applications (30 percent, according to the register) are added to the scan's selection, so staff do not know which cases the scan flagged. UWV carried out a DPIA and an ethical impact assessment. In use since August 2022; no outcome figures are published.","production",[27],[293],[],[316],{"url":317,"title":318,"publisher":151,"date":319},"https://algoritmes.overheid.nl/nl/algoritme/zb000117/15759870/risicoscan-verwijtbare-werkloosheid","Risicoscan Verwijtbare Werkloosheid, Algoritmeregister","2025-12-15",{"level":244,"checkedAt":245},"uwv-unemployment-benefit-risk-scan",{"title":323,"useCases":324,"organization":325,"vendors":327,"summary":328,"stage":329,"year":330,"channels":331,"languages":332,"metrics":333,"outcomeDisclosed":215,"sources":334,"verification":340,"grade":246,"id":341,"organizationSlug":342},"Centers for Medicare & Medicaid Services: anomaly detection on prescription drug cost data",[208],{"name":326,"anonymized":215,"country":254,"region":255,"industry":17},"Centers for Medicare & Medicaid Services",[],"The CMS Division of Payment Reconciliation monitors Prescription Drug Event (PDE) data from Medicare Part D for accuracy. It reported an AI project to identify outliers in that data so that errors can be corrected through outreach to the plans and improper payments, an overpayment or an underpayment of a government benefit, are prevented. The project was at the initiated stage in the 2024 federal inventory; no results are published.","announced",2021,[27],[225],[],[335],{"url":336,"title":337,"publisher":338,"date":339},"https://raw.githubusercontent.com/ombegov/2024-Federal-AI-Use-Case-Inventory/main/data/2024_consolidated_ai_inventory_raw_v2.csv","2024 consolidated AI use case inventory (raw data, version 2)","Office of Management and Budget (GitHub)","2025-01-23",{"level":244,"checkedAt":204},"cms-prescription-drug-cost-anomaly-detection","centers-for-medicare-and-medicaid-services",0,[345],{"kpi":44,"label":346,"unit":229,"aggregate":235,"higherIsBetter":235,"n":347,"nUpTo":343,"median":228,"min":228,"max":228,"byClaimant":348,"vendorOnly":215,"points":349},"Detection improvement",1,{"organization":347,"vendor":343,"regulator":343,"independent":343},[350],{"evidenceId":247,"organization":214,"value":228,"qualifier":230,"claimant":233,"grade":246,"pooled":235},{"low":352,"high":353},500000,13500000,[355,368,397,414,437],{"slug":199,"title":356,"shortTitle":357,"definition":358,"status":9,"industries":359,"functions":360,"patterns":362,"audience":28,"autonomy":29,"adoptionStage":30,"evidenceCount":363,"publicEvidenceCount":363,"organizations":364,"bestGrade":246,"headline":248,"lastVerified":204,"indexable":235},"AI for tax compliance risk scoring and audit selection","Tax compliance risk scoring","Models that score tax returns, taxpayers and transactions for the risk of error, underreporting or fraud, so that a tax administration spends its audit and compliance capacity where the risk is highest, with an officer deciding every compliance action and the selection itself monitored for fairness.",[17],[361,21,19],"risk-management",[23,24],3,[365,366,367],"Belastingdienst","HM Revenue and Customs","Internal Revenue Service",{"slug":200,"title":369,"shortTitle":370,"definition":371,"status":9,"industries":372,"functions":373,"patterns":375,"audience":380,"autonomy":381,"adoptionStage":30,"evidenceCount":382,"publicEvidenceCount":382,"organizations":383,"bestGrade":246,"headline":390,"lastVerified":245,"indexable":235},"AI assistant for benefits eligibility questions and applications","Benefits eligibility and application assistant","An AI assistant that helps people understand which public benefits and grants may apply to them, explains the rules and documents in plain language, guides them through the application and checks it for completeness, while the eligibility decision stays with the agency's rules and caseworkers.",[17],[20,21,374],"customer-service",[376,377,378,379],"conversational-agent","rag-knowledge-assistant","voice-agent","document-processing","customer-facing","copilot",7,[214,384,385,386,387,388,389],"Federal Student Aid (U.S. Department of Education)","Federal Emergency Management Agency","Gemeente Nissewaard","Leeds City Council","Région Provence-Alpes-Côte d'Azur (Région Sud)","YoungWilliams",{"kpi":391,"label":392,"unit":393,"n":394,"nUpTo":343,"kind":395,"value":396,"qualifier":230,"claimant":233,"organization":214,"vendorReported":215},"accuracy","Accuracy","percent",2,"reported",97,{"slug":201,"title":398,"shortTitle":399,"definition":400,"status":9,"industries":401,"functions":402,"patterns":404,"audience":405,"autonomy":29,"adoptionStage":30,"evidenceCount":406,"publicEvidenceCount":406,"organizations":407,"bestGrade":246,"headline":248,"lastVerified":204,"indexable":235},"AI for risk based inspection prioritization in food safety, workplace and environmental regulation","Inspection prioritization","Models that predict which premises, operators or activities are most likely to be non compliant, so that inspectors in food safety, workplace safety, environmental and other regulation spend their visits where the risk is highest, ideally with inspectors choosing the visits and random inspections testing the model.",[17],[361,21,403],"regulatory-compliance",[23,24],"employee-facing",6,[408,409,410,411,412,413],"Care Quality Commission","Driver and Vehicle Standards Agency","U.S. Environmental Protection Agency, Office of Enforcement and Compliance Assurance","Food Standards Agency","Nederlandse Arbeidsinspectie","Nederlandse Voedsel- en Warenautoriteit (NVWA)",{"slug":202,"title":415,"shortTitle":416,"definition":417,"status":9,"industries":418,"functions":423,"patterns":426,"audience":28,"autonomy":428,"adoptionStage":30,"segment":429,"evidenceCount":406,"publicEvidenceCount":406,"organizations":430,"bestGrade":246,"headline":436,"lastVerified":245,"indexable":235},"AI for application and identity fraud detection","Application and identity fraud","AI that checks incoming account and loan applications for forged or AI generated documents, synthetic and stolen identities, and coordinated application rings, by analysing documents, device and application data across the whole queue and cross checking against bureau and official sources.",[419,420,421,17,422],"banking","payments","cross-industry","telecommunications",[19,424,425],"onboarding-and-kyc","lending-and-credit",[379,24,427,23],"computer-vision","supervised-agent","front-office",[431,432,433,214,434,435],"BCU","Close Brothers Motor Finance","CNG Holdings","Payoneer","Telstra",{"kpi":44,"label":346,"unit":229,"n":347,"nUpTo":343,"kind":395,"value":228,"qualifier":230,"claimant":233,"organization":214,"vendorReported":215},{"slug":203,"title":438,"shortTitle":439,"definition":440,"status":9,"industries":441,"functions":444,"patterns":446,"audience":405,"autonomy":381,"adoptionStage":30,"evidenceCount":449,"publicEvidenceCount":449,"organizations":450,"bestGrade":246,"headline":248,"lastVerified":204,"indexable":235},"AI system and model inventory with shadow AI discovery","AI model inventory","A governed register of every AI system and model an organization builds, buys or uses, with its owner, purpose, data, risk tier and approval status, kept current by AI that discovers unregistered use, reads the documentation and assembles the evidence a board, auditor or supervisor asks for.",[421,419,442,17,443],"insurance","manufacturing",[361,403,445],"it-and-engineering",[447,379,377,448],"agentic-workflow","classification-and-routing",4,[451,452,453,454],"Board of Governors of the Federal Reserve System","Office of Management and Budget","Unilever","U.S. Department of Justice",{"indexable":235,"reasons":456},[],[458,464,469,476,482,488,494,501,509,516,523,529,534,541,547,552,559,565,571,577,583,589,595,600,605,612,619,624,629,637,643,649,655,660],{"id":137,"label":459,"issuer":460,"region":147,"url":461,"description":462,"useCases":463,"indexable":235},"EU AI Act","European Union","https://eur-lex.europa.eu/eli/reg/2024/1689/oj","Regulation (EU) 2024/1689: risk based rules for AI systems, with obligations for high risk systems listed in Annex III and transparency duties under Article 50.",197,{"id":138,"label":465,"issuer":460,"region":147,"url":466,"description":467,"useCases":468,"indexable":235},"GDPR","https://eur-lex.europa.eu/eli/reg/2016/679/oj","General Data Protection Regulation, including Article 22 on decisions based solely on automated processing.",180,{"id":140,"label":470,"issuer":471,"region":472,"url":473,"description":474,"useCases":475,"indexable":235},"ISO/IEC 42001","ISO and IEC","global","https://www.iso.org/standard/81230.html","The international management system standard for AI.",110,{"id":139,"label":477,"issuer":478,"region":255,"url":479,"description":480,"useCases":481,"indexable":235},"NIST AI Risk Management Framework","NIST","https://www.nist.gov/itl/ai-risk-management-framework","Voluntary US framework to map, measure, manage and govern AI risk, with a generative AI profile.",83,{"id":483,"label":484,"issuer":460,"region":147,"url":485,"description":486,"useCases":487,"indexable":235},"dora","DORA","https://eur-lex.europa.eu/eli/reg/2022/2554/oj","Digital Operational Resilience Act for financial entities: ICT risk, incident reporting and third party risk, including AI providers.",66,{"id":141,"label":489,"issuer":490,"region":147,"url":491,"description":492,"useCases":493,"indexable":235},"UK GDPR","Information Commissioner's Office","https://ico.org.uk/for-organisations/uk-gdpr-guidance-and-resources/","The UK's version of the GDPR, including rules on solely automated decisions.",64,{"id":495,"label":496,"issuer":497,"region":147,"url":498,"description":499,"useCases":500,"indexable":235},"uk-consumer-duty","FCA Consumer Duty","Financial Conduct Authority","https://www.fca.org.uk/firms/consumer-duty","UK rules that require firms to deliver good outcomes for retail customers, including through automated channels.",47,{"id":502,"label":503,"issuer":504,"region":505,"url":506,"description":507,"useCases":508,"indexable":235},"mas-ai-risk-management","MAS AI risk management guidelines","Monetary Authority of Singapore","asia-pacific","https://www.mas.gov.sg/news/media-releases/2025/mas-guidelines-for-artificial-intelligence-risk-management","Singapore's supervisory expectations for AI risk management at financial institutions, building on the FEAT principles.",36,{"id":510,"label":511,"issuer":512,"region":505,"url":513,"description":514,"useCases":515,"indexable":235},"apra-cps-230","APRA CPS 230","Australian Prudential Regulation Authority","https://www.apra.gov.au/operational-risk-management","Australian operational risk standard covering critical operations and material service providers.",25,{"id":517,"label":518,"issuer":519,"region":472,"url":520,"description":521,"useCases":522,"indexable":235},"pci-dss","PCI DSS","PCI Security Standards Council","https://www.pcisecuritystandards.org/","Security standard for any system that stores, processes or transmits cardholder data.",20,{"id":524,"label":525,"issuer":526,"region":255,"url":527,"description":528,"useCases":522,"indexable":235},"us-sr-11-7","SR 11-7 model risk management","Federal Reserve and OCC","https://www.federalreserve.gov/boarddocs/srletters/2011/sr1107.htm","US supervisory guidance on model risk management, applied by banks to AI and machine learning models.",{"id":142,"label":530,"issuer":531,"region":147,"url":148,"description":532,"useCases":533,"indexable":235},"UK Algorithmic Transparency Recording Standard","UK Government","Mandatory transparency records for algorithmic tools used by UK central government.",16,{"id":535,"label":536,"issuer":537,"region":472,"url":538,"description":539,"useCases":540,"indexable":235},"fatf-recommendations","FATF Recommendations","Financial Action Task Force","https://www.fatf-gafi.org/en/publications/Fatfrecommendations/Fatf-recommendations.html","Global standards for anti money laundering and counter terrorist financing that national rules implement.",15,{"id":542,"label":543,"issuer":460,"region":147,"url":544,"description":545,"useCases":546,"indexable":235},"eu-amlr","EU Anti Money Laundering Regulation","https://eur-lex.europa.eu/eli/reg/2024/1624/oj","Regulation (EU) 2024/1624: the single EU rulebook for customer due diligence, beneficial ownership and suspicious transaction reporting.",14,{"id":548,"label":549,"issuer":460,"region":147,"url":550,"description":551,"useCases":546,"indexable":235},"nis2","NIS2 Directive","https://eur-lex.europa.eu/eli/dir/2022/2555/oj","Directive (EU) 2022/2555 on cybersecurity for essential and important entities, including telecom networks, energy and public administration.",{"id":553,"label":554,"issuer":555,"region":255,"url":556,"description":557,"useCases":558,"indexable":235},"us-bsa","Bank Secrecy Act","FinCEN","https://www.fincen.gov/resources/statutes-and-regulations/bank-secrecy-act","US anti money laundering law: customer due diligence, suspicious activity reports and record keeping.",13,{"id":560,"label":561,"issuer":460,"region":147,"url":562,"description":563,"useCases":564,"indexable":235},"eu-accessibility-act","European Accessibility Act","https://eur-lex.europa.eu/eli/dir/2019/882/oj","Directive (EU) 2019/882: accessibility requirements for banking services, ecommerce and other digital services, applicable since June 2025.",12,{"id":566,"label":567,"issuer":568,"region":255,"url":569,"description":570,"useCases":564,"indexable":235},"hipaa","HIPAA","US Department of Health and Human Services","https://www.hhs.gov/hipaa/index.html","US rules for the privacy and security of protected health information.",{"id":572,"label":573,"issuer":574,"region":472,"url":575,"description":576,"useCases":564,"indexable":235},"telecom-consumer-rules","Telecom consumer protection rules","National telecom regulators","https://www.berec.europa.eu/","National rules on telecom contracts, switching, billing disputes and marketing consent.",{"id":578,"label":579,"issuer":460,"region":147,"url":580,"description":581,"useCases":582,"indexable":235},"eecc","European Electronic Communications Code","https://eur-lex.europa.eu/eli/dir/2018/1972/oj","Directive (EU) 2018/1972: consumer protection, contract, switching and security rules for telecom operators.",11,{"id":584,"label":585,"issuer":586,"region":255,"url":587,"description":588,"useCases":582,"indexable":235},"us-tcpa","Telephone Consumer Protection Act","Federal Communications Commission","https://www.fcc.gov/consumers/guides/stop-unwanted-robocalls-and-texts","US consent rules for automated and prerecorded calls and texts; the FCC has confirmed AI generated voices count as artificial voices.",{"id":590,"label":591,"issuer":504,"region":505,"url":592,"description":593,"useCases":594,"indexable":235},"mas-notice-626","MAS Notice 626","https://www.mas.gov.sg/regulation/notices/notice-626","Singapore's anti money laundering and counter terrorism financing requirements for banks.",10,{"id":596,"label":597,"issuer":460,"region":147,"url":598,"description":599,"useCases":594,"indexable":235},"mifid-ii","MiFID II","https://eur-lex.europa.eu/eli/dir/2014/65/oj","Directive 2014/65/EU on markets in financial instruments: suitability and appropriateness of advice, record keeping and product governance.",{"id":601,"label":602,"issuer":460,"region":147,"url":603,"description":604,"useCases":594,"indexable":235},"eu-psd2","PSD2","https://eur-lex.europa.eu/eli/dir/2015/2366/oj","Payment Services Directive 2: strong customer authentication, transaction risk analysis exemptions and open banking access.",{"id":606,"label":607,"issuer":608,"region":147,"url":609,"description":610,"useCases":611,"indexable":235},"eba-loan-origination","EBA Guidelines on loan origination and monitoring","European Banking Authority","https://www.eba.europa.eu/regulation-and-policy/credit-risk/guidelines-on-loan-origination-and-monitoring","Expectations for credit decisioning, including the use of automated models.",9,{"id":613,"label":614,"issuer":615,"region":255,"url":616,"description":617,"useCases":618,"indexable":235},"us-ecoa-reg-b","ECOA and Regulation B","Consumer Financial Protection Bureau","https://www.consumerfinance.gov/rules-policy/regulations/1002/9/","US fair lending rules, including specific reasons in adverse action notices, which also apply when credit decisions use AI models.",8,{"id":620,"label":621,"issuer":460,"region":147,"url":622,"description":623,"useCases":618,"indexable":235},"solvency-ii","Solvency II","https://eur-lex.europa.eu/eli/dir/2009/138/oj","Directive 2009/138/EC: risk based capital, governance and model requirements for insurers.",{"id":625,"label":626,"issuer":460,"region":147,"url":627,"description":628,"useCases":406,"indexable":235},"eu-idd","Insurance Distribution Directive","https://eur-lex.europa.eu/eli/dir/2016/97/oj","Directive (EU) 2016/97: conduct rules for selling insurance, including demands and needs testing and advice.",{"id":630,"label":631,"issuer":632,"region":633,"url":634,"description":635,"useCases":636,"indexable":235},"cbuae-ai-guidance","CBUAE guidance on AI and ML","Central Bank of the UAE","middle-east","https://www.centralbank.ae/","UAE central bank expectations for the enabling technologies, AI and machine learning used by licensed financial institutions.",5,{"id":638,"label":639,"issuer":640,"region":147,"url":641,"description":642,"useCases":449,"indexable":235},"pra-ss1-23","PRA SS1/23 model risk management","Prudential Regulation Authority","https://www.bankofengland.co.uk/prudential-regulation/publication/2023/may/model-risk-management-principles-for-banks-ss","UK model risk management principles for banks, covering AI and machine learning models.",{"id":644,"label":645,"issuer":646,"region":147,"url":647,"description":648,"useCases":449,"indexable":235},"uk-psr-app-reimbursement","UK APP scam reimbursement rules","Payment Systems Regulator","https://www.psr.org.uk/our-work/app-scams/","Mandatory reimbursement of authorised push payment scam victims by UK payment firms, which shifts scam losses onto banks.",{"id":650,"label":651,"issuer":652,"region":505,"url":653,"description":654,"useCases":363,"indexable":235},"au-scams-prevention-framework","Australian Scams Prevention Framework","Australian Treasury","https://treasury.gov.au/consultation/c2024-573813","Economy wide obligations for banks, telcos and digital platforms to prevent, detect, disrupt and respond to scams.",{"id":656,"label":657,"issuer":460,"region":147,"url":658,"description":659,"useCases":363,"indexable":235},"eu-mar","EU Market Abuse Regulation","https://eur-lex.europa.eu/eli/reg/2014/596/oj","Regulation (EU) 596/2014: insider dealing and market manipulation, including the duty to detect and report suspicious orders and transactions.",{"id":661,"label":662,"issuer":663,"region":255,"url":664,"description":665,"useCases":363,"indexable":235},"us-fcra","Fair Credit Reporting Act","Federal Trade Commission","https://www.ftc.gov/legal-library/browse/statutes/fair-credit-reporting-act","US rules on consumer reports, their accuracy and permissible use, relevant to credit scoring and screening.",1790598299211]