From the Public Safety edition of October 4, 2026
Dallas police seek AI platform to search terabyte-scale case evidence, link offenses and mine cold cases
Dallas Police Department (reported by LAWIN.news) · Municipal policing / investigations · Dallas, Texas, United States
- Publisher
- LAWIN.news
- Original publication
- October 3, 2026
- Source retrieved
- 2026-10-05
- Event date
- 2026-09-17
What happened
LAWIN.news reported on Oct. 3, 2026 that the Dallas Police Department issued a request for information (RFI) for AI tools to help detectives analyze large volumes of digital evidence, link offenses to identify suspects, and review decades-old cold-case files. Maj. Nathan Swyers said on a vendor call that the 2020 Mo3 homicide involved more than a terabyte of video and that the goal is to avoid detectives reviewing footage frame by frame for 8 to 12 hours. DPD already uses Axon software to draft incident reports, Clearview AI facial recognition and another AI summarization tool. Any new tool must meet the FBI CJIS security standard; the notice is an RFI, not a contract process. The HigherGov listing of RFI BD26-00031278 (posted Sept. 17, responses due Oct. 23) describes requested capabilities: transcription of audio and jail calls, video/image analysis, face detection across images, natural-language Q&A over case evidence, automatic linking, Axon Evidence.com integration, 5 TB/3 million-file case support, and compliance with CJIS v6.0 and Texas H.B. 149.
Why it matters
Direct policing/investigations use case: generative and computer-vision AI applied to criminal case evidence, jail calls and cold cases, with procurement language citing CJIS and Texas AI governance law.
Evidence and measured results
Independent local reporting quoting DPD's intelligence-unit commander and a University of Miami criminologist; procurement details cross-checked on the HigherGov aggregator listing (an aggregator-written summary of city documents, not the city portal itself). No performance data exists yet; this is a market-sounding stage.
Limitations and uncertainty
RFI only, with no award, vendor, cost or outcome; LAWIN.news appears to aggregate local reporting (photo credited to KENS 5); HigherGov's description is an aggregator summary and the full RFI document on the city's Bonfire portal was not opened; efficiency and cost-saving rationales are department claims.
Put this evidence to work
Lighthouse Advisory interpretation, grounded in this source. Enriched 2026-10-05; this does not change the original publication date. Each role follows the same checklist so takeaways compare across sources.
Sales
Role takeaway
- Customer problem
- Detectives face terabyte-scale video and mixed digital evidence that takes 8–12 hours of manual frame-by-frame review per pass, plus unreviewed cold-case files.
- Stakeholders
- DPD intelligence unit and investigations command
- Dallas procurement services
- City attorney
- Dallas County District Attorney
- Public defenders
- City council public safety committee
- Discovery questions
- Which case types and evidence volumes would a pilot target first?
- How will AI-generated case links be documented for discovery and Brady disclosure?
- What H.B. 149 documentation and risk assessment does the city require from vendors?
- Will jail-call transcription be in scope, and under what legal authority and retention?
- Value hypothesis
- Bounded: AI-assisted search and transcription may cut evidence review time on large-video cases, measured on closed cases against detective baselines.
- Potential engagement
- RFI-response support and a closed-case evaluation protocol defining metrics, audit logging and DA disclosure rules before any solicitation.
- Unsupported claims
- Improved clearance or cold-case solve rates
- Guaranteed labor or cost savings
- Accuracy of face detection or automatic case linking
- That any vendor has been selected or funded
Pre-sales engineering
Role takeaway
- Fit
- Agencies with large Axon Evidence.com holdings that want AI search, transcription and linking without replacing their evidence system.
- Architecture and integration
- Evidence.com ingest to CJIS-bounded model hosting (transcription, vision, language) to unified index to natural-language Q&A with exhibit citations to audit log export.
- Prerequisites
- Evidence.com API and case-permission mapping
- CJIS v6.0-compliant hosting with no external data sharing
- Labeled closed-case corpus for testing
- H.B. 149 documentation package
- Constraints
- RFI stage only; no funded contract
- Texas H.B. 149 AI-governance requirements
- Jail-call and face-analysis legal limits
- Case sizes up to 5 TB and 3 million files
- Security
- Models must stay inside the agency boundary; every query, answer and linked entity needs immutable logging and role-based access by case.
- Proof of value
- Clip-retrieval recall versus detective review on closed cases
- False-link rate for automatic offense linking
- Transcription word and speaker accuracy on jail calls
- Citation traceability from each answer to source exhibit
Delivery
Role takeaway
- Work
- Design a closed-case evaluation, audit-logging standard and lead-documentation SOP ahead of any solicitation.
- Dependencies
- Axon Evidence.com integration access
- DA and city attorney agreement on disclosure
- CJIS-compliant hosting
- Detective time for ground-truth labeling
- Ownership
- DPD intelligence/investigations unit, with city attorney and DA co-owning disclosure rules.
- Skills and adoption
- Detectives need training on querying, reading citations and corroborating AI-surfaced leads before acting.
- Governance checkpoints
- H.B. 149 documentation review
- DA sign-off on lead documentation
- Privacy review for jail-call and face features
- Council briefing before contract award
- Acceptance criteria
- Proposed: measured review-time reduction on a defined closed-case set
- Proposed: every AI answer cites a source exhibit and timestamp
- Proposed: documented false-link rate below an agreed threshold
- Proposed: complete audit log export for each case
- Risks
- Automation bias treating links as probable cause
- Scope creep into broad jail-call surveillance
- Discovery disputes over AI-generated leads
- Vendor lock-in to the evidence platform
Implementation considerations
Lighthouse Advisory interpretation across the operating dimensions a public-sector buyer must settle before this evidence becomes a design. Each note answers the question under its heading for this specific source.
Architecture and integration
What must connect, and where does the AI sit in the workflow?
The requested design is an evidence-plane AI layer: transcription, vision and language models running over Axon Evidence.com holdings with unified search, entity linking and natural-language question answering, sized for multi-terabyte cases and hosted without external data sharing.
Governance
Who approves, reviews and stays accountable for outcomes?
Natural-language Q&A and automatic case linking over evidence create investigative leads whose provenance must be traceable for discovery and Brady obligations; Texas H.B. 149 documentation is named as a bidder requirement, making state AI-governance law a procurement gate rather than an afterthought.
Security and privacy
What data, permissions and controls need testing?
Jail-call transcription and cross-image face detection extend AI analysis to people who are not suspects; CJIS v6.0 compliance, data containment and audit logs are stated requirements, but retention and model-training restrictions are not described in the reporting.
Accessibility and workforce
Who is affected, and what skills or accommodations follow?
Framed as relief for detective workload amid budget pressure; detectives' role shifts toward validating AI-surfaced leads, which requires new skills rather than fewer reviewers.
Procurement
What should contracts, pricing and exit terms secure?
An RFI that names CJIS v6.0, Texas H.B. 149 documentation, data containment and Evidence.com interoperability gives a template for AI evidence-analysis solicitations; responses are due Oct. 23, 2026.
Operating model
Which teams own the service once it runs?
Evidence review moves from linear human viewing to query-driven triage, so investigative units need lead-documentation and corroboration rules before deployment.
Publication history
- 2026-10-04Public Safety · Issue 274 resources
Stable resource ID: dallas-pd-ai-evidence-rfi-20261003