{"resourceId":"kut-hays-draft-one-da-gate-20260930","versions":[{"version":"external-4fd83496362ea557af8d2ca351331f9bc40f090c2e9d2e37afd2ac273df55988","resource":{"id":"kut-hays-draft-one-da-gate-20260930","title":"KUT: Hays County ends Draft One trial after mixed time savings as Kyle PD withholds tool at DA request","organization":"KUT / Austin NPR (Leigh Walden)","sector":"Law enforcement / prosecution","geography":"United States — Hays County and Kyle, Texas (Austin area)","publishedAt":"30 September 2026","publicationDate":"2026-09-30","eventDate":null,"sourceName":"KUT Radio","sourceLabel":"Independent public-radio reporting with on-record deputies, DA, and ACLU","sourceUrl":"https://www.kut.org/crime-justice/2026-09-30/police-ai-hays-county-san-marcos-kyle-district-attorney","evidenceClass":"independent-reporting","outcomeClass":"mixed","topics":["knowledge-work","governance-procurement","accessibility-workforce","operating-model"],"finding":"Same-day (2026-09-30) KUT reporting: Hays County Sheriff’s Office ran a three-month Axon Draft One trial (BWC audio → transcript → ChatGPT-class narrative). Deputies described faster report return-to-service—Henry cited 20–30 minutes saved on straightforward calls—and intentional “Easter egg” hallucinations to force review; Garcia described planted wizard/superpowers sentences. A $4.5M/10-year expanded Axon contract was unpopular; Chief Deputy Bailey said the county would not seek Draft One funding while still wanting other Axon AI (live translation, unlimited transcription). First Assistant DA Gregg Cox warned AI first drafts may weaken officers’ retained observations and prefers officer-authored detail for charging/plea/trial. Kyle PD already uses some Axon AI but, at DA request, does not use Draft One for reports. ACLU’s Jay Stanley cites dual evidence sets (officer perception vs camera) and accountability uses of reports.","sledRelevance":"Fresh same-day measured-local pilot outcome for U.S. SLED: operational time-savings claims collide with prosecutor non-acceptance and commissioners’ refusal to fund—directly actionable for agencies shopping generative report tools.","evidence":"Opened KUT HTTPS article via curl (HTTP 200); confirmed three-month trial, $4.5M/10-year figure, Henry 20–30 minute claim, Cox and Stanley quotes, Kyle DA gate, and Bailey non-funding statement.","architectureImplications":"Keep Draft One (or equivalents) behind DA-written acceptance rules; log prompts/drafts/edits; do not auto-file AI narratives into RMS without officer attestation fields.","governanceImplications":"Treat prosecutor acceptance and Brady/disclosure readiness as go/no-go gates before purchase; publish intentional-error and review SOPs.","securityPrivacyImplications":"BWC audio leaving the device for cloud transcription/LLM drafting expands CJIS and discovery exposure; retain versioned drafts for disclosure.","caveats":"Single-market radio feature, not a controlled trial; time-savings are deputy self-report; Axon CEO quotes are vendor promotional; hallucination examples include out-of-jurisdiction Utah anecdote.","streamIds":["public-safety"],"roles":{"sales":"Sales should open on prosecutor acceptance, not officer time savings. Hays shows a completed Draft One trial with deputy-reported 20–30 minute gains on simple calls yet no funding ask for continued use, while Kyle already blocks Draft One at the DA’s request. Position any generative report engagement as a bounded acceptance-design workshop with the sheriff, county attorney/DA, and records lead—documenting charging-office rules before quoting Axon SKUs. Do not claim case-quality improvement or guaranteed overtime reduction from this article alone.","engineering":"Engineering fit is optional narrative drafting after BWC upload/transcription with mandatory officer edit and intentional-error scrub—not autonomous report filing. Integrate only if DA policy allows AI-assisted narratives; otherwise park Draft One and evaluate narrower Axon AI (translation/transcription) separately. Prerequisites: CJIS-aligned audio path, editable draft retention, attestation UI, and prompt/version logging. Constraints: ChatGPT-class hallucination risk, memory-displacement concerns from the DA, and dual evidence (camera vs human senses) that architecture cannot reconcile automatically. Proof of value: DA written acceptance, edit-rate telemetry, and zero AI-only filed reports in audit.","delivery":"Delivery work is a time-boxed pilot with prosecutor policy first: SOPs for Easter-egg review, discovery export of drafts/prompts, training against automation bias, and a stop-use trigger if the DA withholds acceptance. Ownership sits with records/patrol command for operations and the DA for charging standards. Acceptance criteria should include documented Kyle-style DA gate decision, measured edit/override rates, and commissioner-ready cost alternatives if Draft One is declined—as Hays did after the three-month trial."},"roleDetails":{"sales":{"customerProblem":"Agencies want faster BWC-to-report drafting but risk DA non-acceptance and political rejection after pilots.","stakeholders":["Sheriff / chief","County commissioners","District attorney / first assistant","Records manager","Patrol deputies","County attorney"],"discoveryQuestions":["Has the DA issued written rules on AI-drafted narratives?","What edit/attestation fields will RMS require?","Will commissioners fund post-trial licenses?","How will drafts/prompts be produced in discovery?"],"valueHypothesis":"Bounded value is optional drafting with prosecutor-approved review controls—not proven charging quality or guaranteed time savings.","potentialEngagement":"4–8 week acceptance-design + limited pilot with DA policy memo before any multi-year Axon AI expansion.","unsupportedClaims":["Guaranteed 20–30 minutes saved on all call types","That Easter eggs catch all hallucinations","Court-ready accuracy from vendor demos"]},"engineering":{"fit":"Officer-in-the-loop narrative assist after BWC transcription; not autonomous report generation.","architectureIntegration":"BWC LTE upload → ASR transcript → LLM narrative draft → officer edit/attest → RMS; retain draft versions for discovery.","prerequisites":["DA acceptance policy","CJIS cloud path","Attestation UI","Prompt/draft retention"],"constraints":["Hallucination risk","Prosecutor memory-displacement concern","Commissioner funding veto after trial"],"security":"Encrypt BWC audio/transcripts; restrict vendor training reuse; log access for Brady packages.","proofOfValue":["Written DA acceptance","Edit-rate and Easter-egg catch metrics","No AI-only filings in sample audit"]},"delivery":{"work":"Stand up SOP, training, discovery export, and go/no-go with DA and commissioners after pilot metrics.","dependencies":["Axon Draft One license","DA policy","RMS fields","Training time"],"ownership":"Patrol/records command owns use; DA owns charging acceptance; county IT owns CJIS path.","skillsAdoption":"Train deputies that drafts are unverified and must be rewritten from independent recall plus video.","governanceCheckpoints":["Pre-pilot DA memo","Mid-trial edit audit","Funding decision with public record"],"acceptanceCriteria":["DA written gate decision recorded","100% drafts officer-attested","Discovery export tested on sample case"],"risks":["Automation bias","Brady gaps on prompts","Political rejection after sunk pilot cost"]}},"retrievedAt":"2026-10-01T02:50:12Z","enrichedAt":"2026-10-01T03:05:00Z","enrichmentBasis":"retrieved source","accessibilityWorkforceImplications":"Time savings on simple calls may not offset review labor; understaffed agencies still need trained editors and DA liaison capacity.","procurementImplications":"Separate Draft One line items from camera/evidence bundles; require DA acceptance and exit rights before multi-year AI add-ons (Hays $4.5M context).","operatingModelImplications":"Make prosecutor policy the operating constraint; pilots without DA buy-in are likely to end as non-funded trials.","sourceVerification":{"openedUrl":"https://www.kut.org/crime-justice/2026-09-30/police-ai-hays-county-san-marcos-kyle-district-attorney","referenceExcerpt":"at the request of the District Attorney’s office, it doesn’t use Draft One in writing their reports.","promptVersion":"sled-research-v4.0","model":null,"basis":"agent-reported inspection"}}}]}