{"resourceId":"uk-local-authority-ai-disclosure-study-2026","versions":[{"version":"external-03238b0cd49931a25b24891035d19d38f5b420f3088656833e9028b11538cd0f","resource":{"id":"uk-local-authority-ai-disclosure-study-2026","title":"New UK council disclosure study exposes an accountability evidence gap, with weak coding reliability","organization":"Governance AI Ltd","sector":"Local-government public accountability","geography":"United Kingdom; limited transfer to U.S. municipal disclosure practice","publishedAt":"Exact publication date unknown; research run dated September 4, 2026","publicationDate":null,"eventDate":"2026-09-04","sourceName":"Governance AI","sourceLabel":"Consultancy-authored disclosure research, not an official government audit","sourceUrl":"https://governanceai.io/research/ai-disclosure-audit-2026","evidenceClass":"independent-research","outcomeClass":"cautionary","topics":["data-security","governance-procurement","accessibility-workforce","operating-model"],"finding":"A consultancy’s document study reports limited visible AI ownership in council reporting, but expressly warns that council coding was its least reliable sector.","sledRelevance":"Interpretation: U.S. localities can use the disclosure questions to review their own public accountability trail. UK percentages and legal references should not be transferred as U.S. prevalence or requirements.","evidence":"Thirty large councils were scored. Published figures report 40% mentioning AI and 3.3% naming an owner. Of four double-coded councils, coders disagreed on three. Non-disclosure does not establish absent controls.","architectureImplications":"Interpretation: Link public inventory entries to internal owners, approvals and change records; publish only information appropriate for public access.","governanceImplications":"Interpretation: Validate a sample manually with department owners before declaring a policy or oversight gap.","securityPrivacyImplications":"Interpretation: Separate publishable oversight evidence from sensitive technical configurations and personal information.","caveats":"Commercial governance-services publisher, nonrepresentative sample, weak local-authority repeatability and anonymized public evidence. Named underlying quotes require a request and were not independently inspected. No operational effectiveness evaluation.","streamIds":["local-government"],"roles":{"sales":"Interpretation: Work with the city manager, clerk, audit committee, CIO and communications lead on whether residents can identify who is accountable for AI. Ask what documents are public, when they were updated and who can verify their completeness. A defensible value hypothesis is clearer accountability and fewer unsupported assurances. Offer a small document-to-owner reconciliation, not a league-table comparison. The study’s council scoring weakness makes it unsuitable for claiming a prospect lacks governance. Avoid compliance or risk-reduction promises. Confirm that a disclosure problem actually exists before proposing a larger assessment or remediation engagement.","engineering":"Interpretation: A simple evidence register may fit better than a new AI platform. Map each public statement to an internal source, owner, effective date and review record. Prerequisites are authoritative policies, an inventory and access to knowledgeable reviewers. Use document search only as assistance; negative matches need human confirmation. Test version drift and conflicting disclosures, and redact sensitive system details. Proposed validation: have two reviewers independently trace a small sample from public text to current controls, reconcile differences and record ambiguity. This workflow assesses evidence traceability; it cannot establish system safety or infer actual practice from disclosure alone.","delivery":"Interpretation: The clerk or governance lead should coordinate document collection, department confirmation, plain-language publication and an update calendar. Dependencies include records access, security review and staff able to resolve conflicting statements. Train contributors to distinguish missing disclosure from missing control. Before publication, require owner approval and accessible formatting. Proposed acceptance: every selected AI service has a confirmed accountable owner, a dated public explanation and an internal supporting record, with all reviewer disagreements resolved or visibly qualified. Recheck after material changes. Risks include stale statements, sensitive over-disclosure and pressure to improve scores without improving accountability."},"retrievedAt":"2026-09-07T03:06:12Z","enrichedAt":"2026-09-07T03:06:12Z","enrichmentBasis":"retrieved source","accessibilityWorkforceImplications":"Interpretation: Make governance records searchable and accessible; teach reviewers that missing keywords are not evidence of absent controls.","procurementImplications":"Interpretation: Require transparent methods and reproducible evidence from governance assessors; do not purchase remediation solely from an opaque score.","operatingModelImplications":"Interpretation: Assign maintenance of public disclosures to a named owner with periodic department confirmation.","sourceVerification":{"openedUrl":"https://governanceai.io/research/ai-disclosure-audit-2026","referenceExcerpt":"Four local authorities were double-coded","promptVersion":"sled-research-v3.1","model":null,"basis":"agent-reported inspection"}}}]}