{"resourceId":"louisiana-fifth-circuit-on-premises-open-weight-ai-court-20261020","versions":[{"version":"external-05dd78ed5bf710de46d84bd8022cf730fd54fd5db8957490aef0f10201530b1c","resource":{"id":"louisiana-fifth-circuit-on-premises-open-weight-ai-court-20261020","title":"Louisiana Fifth Circuit Court of Appeal buys two on-premises AI servers to build open-weight-model tools for filing compliance and intake, kept separate from its case management system","organization":"Louisiana Fifth Circuit Court of Appeal (with ALEA Institute)","sector":"Courts / appellate court operations","geography":"Louisiana, United States","publishedAt":"October 20, 2026 (Judge Scott Schlegel’s newsletter, co-authored with Chief Judge Susan Chehardy)","publicationDate":"2026-10-20","eventDate":null,"sourceName":"[sch]Legal Tech (Judge Scott Schlegel)","sourceLabel":"Building AI Inside the Courthouse","sourceUrl":"https://judgeschlegel.substack.com/p/building-ai-inside-the-courthouse","evidenceClass":"government-evaluation","outcomeClass":"emerging","topics":["infrastructure","developers-agents","governance-procurement","accessibility-workforce"],"finding":"Two judges of the Louisiana Fifth Circuit, including Chief Judge Susan Chehardy, describe the court’s purchase of two AI servers plus added power and cooling, and a partnership with the ALEA Institute to build applications on an open-weight model. Planned first uses are filing compliance checks (page limits, pagination, missing exhibits) and intake classification, later citation verification, prompt-injection screening and eventually neutral bench memoranda. Staff were interviewed about workflows first. The AI environment is kept separate from the case management system, outputs are suggestions that employees verify, and the court continues to use commercial GenAI tools. No results are reported yet.","sledRelevance":"A rare first-hand account of a U.S. state court choosing on-premises open-weight infrastructure alongside commercial cloud tools, with explicit reasoning about cost uncertainty, security responsibilities and isolating experiments from official records. It is directly relevant to court IT leaders weighing hybrid AI architecture.","evidence":"Opened the Oct. 20 post co-authored by the chief judge. It is an operator self-account describing purchases, design choices and plans; it reports no costs, server specifications, model names, performance data or timelines. Future capabilities such as citation verification and bench memoranda are stated as goals.","architectureImplications":"Hybrid model: on-premises open-weight inference for court-built tools, commercial cloud GenAI retained; the experimental environment is isolated from the case management system of record.","governanceImplications":"Outputs are advisory; employees perform official filing review; the system will not recommend rulings; closer integration requires rigorous testing first.","securityPrivacyImplications":"The judges note on-premises does not make a system immune to security issues; planned screening for prompt injection in incoming filings.","caveats":"Self-reported account by the court’s own judges; early stage; no metrics or costs; partner named with a non-endorsement disclaimer; applicability to courts without IT capacity is limited.","streamIds":["public-safety"],"roles":{"sales":"Sales framing: a state appellate court decided to own AI servers while keeping commercial tools, because costs and capabilities are uncertain. The bounded opportunity is helping similar courts compare on-premises and cloud options for narrow tasks such as filing checks. This source does not show cost savings, accuracy or staff time saved, and it explicitly says not every court needs its own servers.","engineering":"Engineering framing: this is a hybrid architecture with on-premises open-weight inference, cloud GenAI in parallel, and a hard boundary from the case management system. Pre-sales should assess power, cooling, patching and security operations, and which tasks a smaller model can perform reliably. Proof should compare page-limit, pagination and missing-exhibit checks against clerk findings on historical filings.","delivery":"Delivery framing: the work begins with workflow interviews, then narrow advisory tools that staff verify, with no writes to official records. Dependencies are facility upgrades, a development partner and in-house IT capacity. The owner is the clerk of court with court IT. Acceptance should be judged on detection accuracy and staff verification effort before any integration with the case system."},"roleDetails":{"sales":{"customerProblem":"Courts face rising workloads and uncertain AI costs, and want tools fitted to their own filing rules.","stakeholders":["Chief judge","Clerk of court","Court IT director","Staff attorneys","Self-represented litigant services"],"discoveryQuestions":["Which filing defects cause the most rework?","Does the court have facility capacity for AI hardware?","What commercial GenAI tools are already in use?","Who would verify AI suggestions?"],"valueHypothesis":"Narrow, rule-based filing checks on court-controlled infrastructure may reduce avoidable deficiencies if validated.","potentialEngagement":"Options analysis comparing on-premises and cloud deployment for one filing-compliance use case.","unsupportedClaims":["Cost savings versus cloud","Accuracy of compliance checks","That bench memoranda tools are ready"]},"engineering":{"fit":"Courts with in-house IT and facility capacity considering hybrid AI architecture.","architectureIntegration":"Isolated on-premises inference environment reading copies of filings, returning advisory flags, with no write access to the case management system.","prerequisites":["Server room power and cooling","Open-weight model selection","Documented filing rules and staff workflows"],"constraints":["Hardware operations burden","Limited in-house AI expertise","Local court rules vary"],"security":"Patch and monitor AI servers like other critical systems and screen incoming documents for prompt-injection content.","proofOfValue":["Proposed: run compliance checks on past filings and compare with clerk-identified defects","Proposed: measure staff time to verify suggestions"]},"delivery":{"work":"Interview staff, build advisory filing-check tools, and evaluate them in isolation before integration.","dependencies":["Facility upgrades","Development partner","IT staff availability"],"ownership":"Clerk of court and court IT director under the chief judge","skillsAdoption":"IT staff learn model operations alongside the partner; clerks learn to verify and override flags.","governanceCheckpoints":["No writes to official records","Testing gate before any CMS integration","Separate validation for each new task type"],"acceptanceCriteria":["Proposed: defect detection precision and recall measured on historical filings","Proposed: zero automated changes to the case record"],"risks":["Security assumed rather than verified","Scope creep toward judicial drafting","Key-person dependency on partner"]}},"retrievedAt":"2026-10-21T02:51:04Z","enrichedAt":"2026-10-21T02:57:27Z","enrichmentBasis":"retrieved source","accessibilityWorkforceImplications":"Pre-submission feedback could help self-represented litigants; staff knowledge was captured through interviews; in-house IT is learning alongside the development partner.","procurementImplications":"Capital purchase of servers and facility upgrades plus a development partnership, instead of buying a ready-made product.","operatingModelImplications":"Court IT must operate power, cooling, maintenance and cybersecurity for AI hardware; tools are built from staff-described workflows outward.","sourceVerification":{"openedUrl":"https://judgeschlegel.substack.com/p/building-ai-inside-the-courthouse","referenceExcerpt":"For now, we are keeping the AI environment separate from our existing case management system.","promptVersion":"sled-research-v4.0","model":null,"basis":"agent-reported inspection"}}}]}