{"resourceId":"berkeley-community-supervision-evidence-review-2026","versions":[{"version":"external-08fd371e793bc814f408339266841a3b9a61c21eb4d15de173a762466ccaa678","resource":{"id":"berkeley-community-supervision-evidence-review-2026","title":"Berkeley review connects supervision technology errors with weak routes to challenge evidence","organization":"UC Berkeley School of Law, Samuelson Clinic","sector":"Courts, corrections and community supervision","geography":"United States; federal courts and five selected states","publishedAt":"February 2026; PDF gives month only","publicationDate":null,"eventDate":null,"sourceName":"UC Berkeley Law","sourceLabel":"Legal and technical white paper; selected-jurisdiction review","sourceUrl":"https://www.law.berkeley.edu/wp-content/uploads/archive/2026/02/2026_02_26_CommunitySupervisionWhitePaper_final.pdf","evidenceClass":"academic-research","outcomeClass":"cautionary","topics":["knowledge-work","developers-agents","data-security","governance-procurement","accessibility-workforce","operating-model"],"finding":"The authors argue that community-supervision procedures can fail to expose unreliable technological evidence, including AI-enabled monitoring.","sledRelevance":"Historical source newly added for U.S. probation, parole and court evidence workflows, complementing the archived corrections project audit without repeating that source.","evidence":"The paper reviews technologies and procedures in federal courts, California, Georgia, Indiana, New York and Texas, informed by attorney interviews. It is not a comparative deployment trial.","architectureImplications":"Interpretation: Preserve original sensor records, model versions and human changes as separate artifacts with controlled disclosure exports.","governanceImplications":"Interpretation: Map locally applicable review and challenge procedures with counsel before using automated assertions in enforcement.","securityPrivacyImplications":"Interpretation: Limit biometric and location collection; separate supportive communications from enforcement evidence and record authorized reuse.","caveats":"No population error estimate or measured reform benefit. Technologies include non-AI devices. The PDF is dated February; its filename and landing-page date do not establish an exact PDF publication date.","streamIds":["public-safety"],"roles":{"sales":"Interpretation: Probation directors, court administrators, defenders and procurement staff may struggle to explain or challenge a device-generated allegation. Ask what evidence is available when a person disputes an alert, how equipment failures are handled and who pays for technical review. A bounded engagement could audit the existing evidence handoff using synthetic cases. The value hypothesis is a more dependable and reviewable process. This legal review does not demonstrate reduced incarceration or savings, and it does not establish a need to purchase additional surveillance.","engineering":"Interpretation: Fit is an evidence-preservation and review layer, not autonomous violation adjudication. Prerequisites include documented device behavior, model versions, access authority and a local procedural map. Test clock drift, network loss, identity mismatches and missing source data through the complete export path. Compare exported records with originals and verify that a reviewer can reconstruct why an alert occurred. Hosting should follow sensitivity and availability requirements; no particular platform is validated here. Keep chatbot output distinct from observations and block automatic conversion into a sanction request.","delivery":"Interpretation: A community-supervision operations owner should coordinate counsel, records staff, support teams and an independent technical reviewer. Inventory devices, document failure handling and train staff to distinguish a machine assertion from corroborated evidence. Proposed acceptance: all synthetic disputed alerts can be traced to their source, authorized reviewers can access required artifacts, and simulated outages route to human support. Review accessibility with affected users before launch. Dependencies include contracts permitting inspection and staffed correction channels. Risks include vendor opacity, delayed disclosure and treating a connectivity failure as noncompliance."},"retrievedAt":"2026-09-10T03:02:08Z","enrichedAt":"2026-09-10T03:03:38Z","enrichmentBasis":"retrieved source","accessibilityWorkforceImplications":"Interpretation: Provide accessible notice, support and alternatives for device or connectivity failures.","procurementImplications":"Interpretation: Require inspectable validation materials and case-level exports before accepting a monitoring tool.","operatingModelImplications":"Interpretation: Supervision operations, defense access coordinators and technical evidence specialists need an explicit handoff process.","sourceVerification":{"openedUrl":"https://www.law.berkeley.edu/wp-content/uploads/archive/2026/02/2026_02_26_CommunitySupervisionWhitePaper_final.pdf","referenceExcerpt":"The analysis also relies on interviews with attorneys who practice parole and probation law.","promptVersion":"sled-research-v3.1","model":null,"basis":"agent-reported inspection"}}}]}