{"resourceId":"uk-anpr-analytics-tool-faculty-ai-national-adoption-tbij-20261106","versions":[{"version":"external-4a0fd7b0036348e5da41e0988988e11a1f370f78f783ec90c8399311423a5eb1","resource":{"id":"uk-anpr-analytics-tool-faculty-ai-national-adoption-tbij-20261106","title":"UK police AI tool that searches every ANPR journey is being upgraded by Faculty AI for \"other crime types\" and \"future national adoption\", with FOI requests refused","organization":"Police Digital Service / National Police Chiefs’ Council / Home Office (reported by The Bureau of Investigative Journalism)","sector":"Policing / automated number plate recognition analytics","geography":"United Kingdom (England and Wales regional organised crime units)","publishedAt":"TBIJ, Nov. 6, 2026 (Mark Wilding); contract award notice published Aug. 13, 2026","publicationDate":"2026-11-06","eventDate":"2026-11-06","sourceName":"The Bureau of Investigative Journalism","sourceLabel":"The UK has its own network of Flock-style AI cameras. So where’s the outrage?","sourceUrl":"https://www.thebureauinvestigates.com/stories/2026-11-06/flock-cameras-us-anpr-surveillance-uk","evidenceClass":"independent-reporting","outcomeClass":"cautionary","topics":["data-security","governance-procurement","operating-model"],"finding":"TBIJ reports that UK police have an AI tool, formerly called ANPR Find and Profile and now the ANPR Analytics Tool, that searches all journeys in the national ANPR system and flags vehicles for investigation. When TBIJ first reported it, three regional organised crime units were trialling it against county lines drug dealing. A recently published contract awards Faculty AI about half a million pounds to develop machine learning so the tool can serve \"other crime types\" with a view to \"future national adoption\". The network has about 12,000 static and nearly 2,000 mobile cameras, captures more than 100 million vehicle movements a day, and records are typically kept for a year. The Home Office says the tool does not profile people or vehicles or make automated decisions; the Home Office, NPCC and forces refused or failed to answer FOI requests, and the Information Commissioner noted uninvolved people may have journeys tracked. The award notice (UKSpending, opened separately) lists £527,497 from Police Digital Service and cites \"significant increases in the success rate\" of county-lines stop and searches in trials, without published figures.","sledRelevance":"U.S. agencies and vendors are moving from plate lookups to pattern analytics across all reads. The UK case shows a national-scale version, its contracting path, and the transparency gap that U.S. councils are now probing.","evidence":"Opened the TBIJ article via HTTPS curl (User-Agent SLED-research-bot/1.0) and saved text; opened the UKSpending award notice for the same contract to confirm buyer, supplier, value and stated purpose.","architectureImplications":"Bulk pattern search over a national read store is a different system from hotlist matching: it needs analytic pipelines, export controls on CSV extracts and query logging at the analyst level.","governanceImplications":"Expansion from one crime type toward national use is happening through procurement rather than a dedicated legal basis; FOI refusals limit external review.","securityPrivacyImplications":"One-year retention and suspicionless pattern search raise the exposure of uninvolved drivers, as the ICO observation notes.","caveats":"Journalism with a stated viewpoint; no tool accuracy, false-positive or stop-outcome data are published; \"significant increases\" in stop success is an unquantified claim in the award notice; UK legal context differs from U.S. Fourth Amendment and state ALPR laws.","streamIds":["public-safety"],"roles":{"sales":"Sales framing: UK police are turning plate cameras from lookup tools into AI pattern search across every journey, and funding national rollout through a modest contract. U.S. buyers considering similar analytics will face the same questions TBIJ raised: what it flags, how often it is wrong, and who can see the logs. A bounded engagement could be an analytics governance review before purchase. Do not claim the tool improves stop outcomes by any measured amount; no figures were published.","engineering":"Engineering view: this is bulk analytics over a national read store, with CSV exports processed into flagged vehicles and intelligence packs. Any comparable U.S. build needs per-analyst query logging, export controls, retention aligned to state law, and a documented model for what counts as unusual. Proof of value should measure precision of flags against confirmed cases and stop outcomes, because the source provides no accuracy data. Treat vendor statements that it does not profile as claims to test, not as properties.","delivery":"Delivery plan: before any pattern-analytics pilot, write a use policy naming permitted crime types, retention, and who may export data; set an evaluation with baseline stop success rates and false-flag review; and publish a summary. Dependencies are records counsel, the data owner and an oversight body. Acceptance should be a published evaluation before scope expands to new crime types. The main risk is quiet scope creep from one offence type to general use without public authorization."},"roleDetails":{"sales":{"customerProblem":"Agencies want more investigative value from plate-reader data but face public backlash over suspicionless search.","stakeholders":["Police intelligence leads","Records and privacy counsel","Oversight board or council","Data protection officer"],"discoveryQuestions":["Do you run pattern queries across all reads or only hotlist and case lookups?","What retention applies to reads used for analytics?","Could you answer an FOI request describing how flags are generated?"],"valueHypothesis":"A governed analytics pilot with published measures may add investigative leads while reducing legitimacy risk, as interpretation.","potentialEngagement":"Four-week governance and evaluation design for ALPR pattern analytics before procurement.","unsupportedClaims":["That the UK tool measurably improves stop outcomes","That it does not profile people, beyond Home Office assertion","That the approach is lawful under U.S. state ALPR statutes"]},"engineering":{"fit":"Relevant to agencies with large ALPR stores considering pattern or anomaly analytics.","architectureIntegration":"Batch export from the read store into an analytics service producing flagged vehicles and intelligence packs for local units.","prerequisites":["Read store with lawful retention","Analyst role definitions","Query and export logging"],"constraints":["State ALPR retention and sharing laws","Public records obligations","No published accuracy benchmarks"],"security":"Restrict CSV exports, log every query by analyst, and audit access by non-police agencies sharing the data.","proofOfValue":["Precision of flags against confirmed cases","Stop outcomes from flagged versus non-flagged leads","Audit trail completeness test"]},"delivery":{"work":"Policy, evaluation design and limited pilot of pattern analytics with published results before expansion.","dependencies":["Data owner","Counsel","Oversight body","Analytics vendor"],"ownership":"Intelligence unit commander with privacy officer sign-off.","skillsAdoption":"Analysts need training on interpreting flags as leads, not grounds, and on documenting follow-up.","governanceCheckpoints":["Use policy approved","Evaluation baseline set","Results published before scope change"],"acceptanceCriteria":["Flag precision reported against an agreed threshold","No unlogged exports in audit sample","Public summary released within pilot period"],"risks":["Scope creep to new crime types","FOI and legitimacy challenges","Unmeasured false flags driving stops"]}},"retrievedAt":"2026-11-08T03:52:26Z","enrichedAt":"2026-11-08T03:58:41Z","enrichmentBasis":"retrieved source","accessibilityWorkforceImplications":"Shifts work toward data analysts in regional units; officer training on acting on flagged vehicles is not described.","procurementImplications":"The award ties funding to evidence for national adoption; buyers elsewhere should require published evaluation criteria and success measures in similar contracts.","operatingModelImplications":"Analysts produce intelligence packs that lead to local stop and search, so downstream stop outcomes and error rates should be tracked.","sourceVerification":{"openedUrl":"https://www.thebureauinvestigates.com/stories/2026-11-06/flock-cameras-us-anpr-surveillance-uk","referenceExcerpt":"a recently published contract reveals that Faculty AI, a London tech company, has since been awarded half a million pounds to upgrade the tool","promptVersion":"sled-research-v4.0","model":null,"basis":"agent-reported inspection"}}}]}