{"resourceId":"poland-supreme-administrative-court-ai-clerk-assistant-pilot-20261020","versions":[{"version":"external-05dd78ed5bf710de46d84bd8022cf730fd54fd5db8957490aef0f10201530b1c","resource":{"id":"poland-supreme-administrative-court-ai-clerk-assistant-pilot-20261020","title":"Poland’s Supreme Administrative Court signs AI cooperation with the Ministry of Digital Affairs: 30 court staff will pilot an AI clerk assistant, with every AI-drafted document approved by a judge","organization":"Supreme Administrative Court of Poland (NSA) and Ministry of Digital Affairs","sector":"Courts / judicial administration","geography":"Poland","publishedAt":"20 October 2026 (NSA news)","publicationDate":"2026-10-20","eventDate":"2026-10-20","sourceName":"Naczelny Sąd Administracyjny","sourceLabel":"AI to support the administrative judiciary. AI assistant for judges – PLLuM at the SAC","sourceUrl":"https://www.nsa.gov.pl/en/news/events/ai-to-support-the-administrative-judiciary-ai-assistant-for-judges-pllum-at-the-sac/","evidenceClass":"government-evaluation","outcomeClass":"emerging","topics":["knowledge-work","governance-procurement","data-security","accessibility-workforce"],"finding":"On Oct. 20 the Supreme Administrative Court and the Ministry of Digital Affairs signed a letter of intent to bring AI into the administrative judiciary. In the first stage, 30 court staff “ambassadors” will test a pilot of the ministry’s AI Clerk Assistant for summarising and comparing documents, drafting letters, searching legal provisions and anonymisation. Every AI-generated document must be checked and approved by a judge. Pilot results will shape a dedicated AI Assistant for Administrative Court Judges, supported by the national PLLuM model. In return, the court will make rulings from its central database available to NASK solely to improve the AI tools.","sledRelevance":"International courts example with limited direct transferability: it pairs a government-built shared assistant with a court-specific follow-on, a small staff pilot cohort, mandatory judicial approval, and a data-sharing exchange in which court rulings train public-sector models. U.S. state court administrators weighing shared state AI platforms face similar questions.","evidence":"Opened the court’s English news item dated 20 October 2026. It is an official announcement of intent; no performance results, evaluation metrics, timeline or budget are given. The Polish ministry release and a CIRE report carry the same content.","architectureImplications":"A central government assistant is reused by the court, with a court-specific tool planned on the same base and a national language model; court rulings flow to the model developer for improvement.","governanceImplications":"Human approval is mandatory for every AI-generated document; the court frames cooperation as voluntary and transparent; the pilot precedes any judge-facing tool.","securityPrivacyImplications":"Anonymisation is one of the tested functions; sharing rulings for model training raises data-protection questions the release does not address.","caveats":"Announcement only; no results, metrics, schedule or costs; administrative courts in a civil-law system; transferability to U.S. trial courts is limited.","streamIds":["public-safety"],"roles":{"sales":"Sales framing: a national court is starting AI with a small staff cohort, a shared government assistant, and a rule that judges approve every AI-drafted document. For U.S. court administrators, the bounded offer is help designing a pilot evaluation, not selling a judge assistant. The source supports discussion of pilot structure and governance only; it cannot support claims about time savings, accuracy or judicial acceptance, because none were reported.","engineering":"Engineering framing: the design reuses a government-wide assistant and a national model, then specialises for court work, while court rulings are shared back for model improvement. Pre-sales should test where court data would be processed, how anonymisation is validated, and whether outputs record the source documents used. Proof should measure summarisation and anonymisation quality on real court files before any judge-facing release.","delivery":"Delivery framing: the work is a staged pilot: recruit staff ambassadors, define evaluation criteria for each function, enforce judicial approval, and decide what moves to a judge tool. Dependencies are the shared platform, data-sharing terms and digitised files. The owner is court administration with judicial oversight. Acceptance should be defined before testing starts, not inferred from adoption counts."},"roleDetails":{"sales":{"customerProblem":"Courts want AI help with summarising, drafting and anonymisation without undermining judicial responsibility.","stakeholders":["Court administrator","Chief judge","Judicial assistants","State IT or digital agency","Data protection officer"],"discoveryQuestions":["Is there a shared state AI platform the court could use?","Which tasks would staff pilot first?","Who approves AI-drafted documents?","Would court data be shared to improve models?"],"valueHypothesis":"A structured staff pilot with defined criteria can show where AI assistance is reliable before judges rely on it.","potentialEngagement":"Pilot design and evaluation framework for a court staff AI assistant across four task types.","unsupportedClaims":["Time or cost savings","Accuracy of summaries or anonymisation","That the judge assistant will be built"]},"engineering":{"fit":"Courts with access to a shared government AI platform and digitised case files.","architectureIntegration":"Shared government assistant accessed by court staff, with a later court-specific tool on the same model base.","prerequisites":["Digitised documents","Defined task list","Data-sharing agreement terms"],"constraints":["Civil-law administrative court context","Data protection rules on ruling reuse","No published technical details"],"security":"Validate anonymisation before any ruling data leaves the court, and log inputs and outputs for review.","proofOfValue":["Proposed: score summaries against staff-written summaries on sample files","Proposed: test anonymisation recall on documents with known personal data"]},"delivery":{"work":"Run a staff ambassador pilot with task-level evaluation and mandatory judicial approval.","dependencies":["Ministry platform availability","Signed data terms","Staff time for evaluation"],"ownership":"Court administration with a designated judicial lead","skillsAdoption":"Training staff ambassadors to evaluate outputs and coach colleagues.","governanceCheckpoints":["Evaluation criteria agreed before pilot","Judicial approval enforced for every AI draft","Review before moving to a judge-facing tool"],"acceptanceCriteria":["Proposed: documented error rates for each pilot function","Proposed: zero AI-generated documents issued without recorded judicial approval"],"risks":["Over-reliance on drafts","Data-sharing objections","Pilot results not published"]}},"retrievedAt":"2026-10-21T02:51:03Z","enrichedAt":"2026-10-21T02:57:27Z","enrichmentBasis":"retrieved source","accessibilityWorkforceImplications":"The minister stresses no job losses; 30 staff ambassadors will evaluate the tool and presumably support adoption.","procurementImplications":"Interpretation: the court obtains tools through an inter-agency arrangement rather than a commercial procurement, trading data access for tooling.","operatingModelImplications":"Pilot first with court staff, then a judge-specific assistant for adjudication support and case-file management.","sourceVerification":{"openedUrl":"https://www.nsa.gov.pl/en/news/events/ai-to-support-the-administrative-judiciary-ai-assistant-for-judges-pllum-at-the-sac/","referenceExcerpt":"Every document generated by AI will have to be checked and approved by a judge.","promptVersion":"sled-research-v4.0","model":null,"basis":"agent-reported inspection"}}}]}