{"resourceId":"new-mexico-ai-fabricated-testimony-sanction-20260911","versions":[{"version":"external-bf25cbbb21744e6f8a5fe021eaaaefa25397ac809c220d49bbccfdf70b48a2a9","resource":{"id":"new-mexico-ai-fabricated-testimony-sanction-20260911","title":"New Mexico appeal exposes fabricated testimony in an unchecked AI-assisted brief","organization":"New Mexico Supreme Court; reporting by Mike Scarcella","sector":"Criminal courts and defense","geography":"New Mexico, United States","publishedAt":"September 11, 2026","publicationDate":"2026-09-11","eventDate":"2026-09-09","sourceName":"Reuters via Investing.com","sourceLabel":"Original Reuters reporting, syndicated full text","sourceUrl":"https://www.investing.com/news/economy-news/chatgpt-invented-fake-police-testimony-in-murder-appeal-new-mexico-high-court-says-4897839","evidenceClass":"independent-reporting","outcomeClass":"cautionary","topics":["knowledge-work","developers-agents","data-security","governance-procurement","accessibility-workforce","operating-model"],"finding":"Reuters reports that New Mexico's Supreme Court sanctioned Stephen Aarons after an appeal brief included invented witnesses and testimony attributed to ChatGPT.","sledRelevance":"A current state criminal-defense failure illustrates the stakes of source verification in drafting assistance.","evidence":"The reported sanction was $5,000 and contempt, with disciplinary referral. Aarons said he used transcripts and case materials for summarization and misunderstood hallucination risk. The appeal remained pending. This is one reported incident, without a comparative sample or error-rate denominator.","architectureImplications":"Interpretation: Keep transcript passages alongside drafts, preserve version history, and prevent a drafting agent from filing.","governanceImplications":"Interpretation: Assign an attorney to verify every factual assertion independently before signature.","securityPrivacyImplications":"Interpretation: Authorize the processing environment before uploading case records; assess retention and training-use terms.","caveats":"Full court order and underlying prompts were not inspected; the accessible case listing was truncated. Attribution relies on Reuters, including the lawyer's response. No model version, prevalence or appeal outcome is established.","streamIds":["public-safety"],"roles":{"sales":"Interpretation: Public defenders, appellate supervisors and court IT teams need to reduce drafting burden without losing control of the record. Ask how transcripts enter the workflow, who checks witness identities, and whether reviewers can see the exact cited passage. A bounded engagement could examine one appeal-preparation process and build a verification exercise using authorized historical material. The value hypothesis is identifying preventable review failures and quantifying the effort required for safe assistance. Avoid promising better appeal results, zero hallucinations or savings from this incident. Suitability depends on local practice and the team's capacity to perform substantive review, including under deadline pressure.","engineering":"Interpretation: Fit is an isolated drafting workspace with document-level permissions and source-linked assertions. Prerequisites include reliable transcripts, stable passage identifiers and an attorney-adjudicated reference set. Test nonexistent names, misattribution, contradictory passages and unsupported quotations using the proposed current model. Compare assisted reviewers with the existing process and measure both missed errors and correction effort. Retrieval can support checking but does not guarantee fidelity. Cloud, local and hybrid choices require separate privacy and operational assessment; this story establishes no product compliance status. Keep generation separate from case-system updates and filing credentials, and log model changes for repeat evaluation.","delivery":"Interpretation: An appellate practice lead should own the process with records staff, training personnel and IT. Map transcription, draft creation, review and signature; establish a conventional baseline and run a reversible shadow pilot. Dependencies include protected review time and lawful access to the record. Train staff to verify facts as well as legal references, and provide accessible source-navigation instructions. Proposed acceptance: every factual assertion in a sampled final draft has a checked source location, and any fabricated witness stops release pending correction. These are proposed controls, not observed results. Track reviewer workload and abandoned drafts; preserve a manual route when validation cannot be completed."},"retrievedAt":"2026-09-13T03:02:23Z","enrichedAt":"2026-09-13T03:05:28Z","enrichmentBasis":"retrieved source","accessibilityWorkforceImplications":"Interpretation: Budget review time and accessible transcript navigation; do not treat fluent prose as staff competence.","procurementImplications":"Interpretation: Require exportable source links, retention controls and evaluation rights before adopting a legal copilot.","operatingModelImplications":"Interpretation: Place release authority with legal staff and maintain correction and escalation procedures.","updateExplanation":"New-to-archive September 11 reporting adds a factual-testimony failure distinct from earlier prosecution experiments and court guidance.","sourceVerification":{"openedUrl":"https://www.investing.com/news/economy-news/chatgpt-invented-fake-police-testimony-in-murder-appeal-new-mexico-high-court-says-4897839","referenceExcerpt":"contained false testimony from wholly fabricated witnesses","promptVersion":"sled-research-v3.1","model":null,"basis":"agent-reported inspection"}}}]}