From the Emergency Services edition of September 20, 2026
explainx: New Orleans Carbyne 911 AI narrow but unaudited—zero FP/FN claim remains unverified
explainx.ai analysis of OPCD / Carbyne (Axon) deployment · 911 AI governance / independent scrutiny · New Orleans, Louisiana, United States (national transferability discussion)
- Publisher
- explainx.ai
- Original publication
- 2026-08-10
- Source retrieved
- 2026-09-21
- Event date
- 2026-08-06
What happened
An Aug 10 2026 explainx deep dive argues New Orleans has run Carbyne AI on live 911 traffic since 2023 for crash-call deduplication (with ~200 m matching and human escalation), confirmed publicly around Aug 6 2026 after limited prior disclosure, and itemizes what remains unconfirmed—including any independent third-party audit of the oft-repeated zero false-positive / zero false-negative claim from the initial fully monitored window. Why now: pairs with Gambit's scope interview to give SLED buyers a scrutiny checklist after non-emergency AVA and CommsCoach themes already in the archive.
Why it matters
Independent scrutiny template for U.S. SLED PSAPs evaluating vendor accuracy claims on live emergency call AI.
Evidence and measured results
Opened explainx HTML dated Aug 10, 2026. TL;DR table lists unconfirmed independent audit of zero FP/FN, full incident-level error rate since 2023, and contract terms beyond a reported ~$600K/year figure. Describes classifier-style architecture, 100% human monitoring at launch, and escalation on uncertainty.
Limitations and uncertainty
Secondary analysis synthesizing press and official statements; not a formal government audit; some dollar and staffing figures are reported, not independently verified here.
Put this evidence to work
Lighthouse Advisory interpretation, grounded in this source. Enriched 2026-09-20; this does not change the original publication date. Each role follows the same checklist so takeaways compare across sources.
Sales
Role takeaway
- Customer problem
- Agencies need to separate narrow, defensible 911 AI designs from unaudited accuracy marketing.
- Stakeholders
- PSAP directors
- City auditors
- Procurement
- Civil liberties advocates
- Vendor account teams
- Discovery questions
- Was the zero-error window 100% monitored?
- Who can independently sample closes?
- Is live use disclosed to callers?
- What monitoring remains after month three?
- Value hypothesis
- If independent sampling confirms low misclose rates under narrow rules, councils may approve continued use without broader automation.
- Potential engagement
- Audit-rights contract annex plus 180-day independent sampling study.
- Unsupported claims
- Permanent zero FP/FN
- Proof that AI can safely answer medical 911
- Transfer of New Orleans rates to other cities without local audit
Pre-sales engineering
Role takeaway
- Fit
- Duplicate-call classifier with mandatory uncertainty escalation and auditable close logs.
- Architecture and integration
- Immutable close/transfer logs tied to recordings; metrics API for external auditors.
- Prerequisites
- Exportable event logs
- Retained audio
- Sampling protocol
- Third-party NDA/access
- Constraints
- Vendor black-box models
- Post-pilot spot-check drift
- Staffing for review
- Security
- Ban cross-customer training on PSAP audio; encrypt recordings at rest.
- Proof of value
- Independent FP/FN sample after monitored launch
- Disclosure completeness check
- Escalation latency distribution
Delivery
Role takeaway
- Work
- Stand up disclosure, retain logs, commission sample audit, brief board on results.
- Dependencies
- Budget for auditor
- Vendor cooperation
- Legal review
- QA staffing
- Ownership
- PSAP director with city auditor co-sponsor.
- Skills and adoption
- Train executives to read pilot vs steady-state accuracy correctly.
- Governance checkpoints
- Kickoff audit
- Midpoint sample
- Expansion vote
- Acceptance criteria
- Independent sample published
- No new use cases until sample accepted
- Caller disclosure live
- Risks
- Vendor refusal of data access
- Public backlash to delayed disclosure
- Silent monitoring reduction
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?
Treat vendor accuracy claims as pilot-window observations unless continuous sampling and third-party audit are contractually required; separate classifier screening from any generative agent stack.
Governance
Who approves, reviews and stays accountable for outcomes?
Mandate public disclosure timelines, independent error audits, and change-control before expanding beyond crash duplicates; document monitoring intensity after the initial 100% review period.
Security and privacy
What data, permissions and controls need testing?
Axon/Carbyne ownership and multi-year call handling raise supply-chain and data-use questions; require written bans on reuse of 911 audio for unrelated model training.
Accessibility and workforce
Who is affected, and what skills or accommodations follow?
Highlights staffing shortfalls as the motive for automation while warning that audit labor is still required; accessibility of disclosure materials for the public is part of trust.
Procurement
What should contracts, pricing and exit terms secure?
Negotiate audit rights, data-export SLAs, and training-use bans up front; treat ~$600K/year class spend as needing performance schedules.
Operating model
Which teams own the service once it runs?
Encourages a two-phase operating model: 100% review at launch, then statistically valid sampling—never marketing the pilot window as permanent accuracy.
Publication history
- 2026-09-20Emergency Services · Issue 155 resources
Stable resource ID: explainx-nola-carbyne-audit-gaps-20260810