From the Emergency Services edition of September 19, 2026
Long Beach publishes CommsCoach privacy report card: 100% QA evals, GovCloud, 24-hour unmapped deletion
City of Long Beach Disaster Preparedness and Emergency Communications (DPEC) · 911 / ECC quality assurance governance · Long Beach, California, United States
- Publisher
- City of Long Beach Smart City / privacy evaluation package
- Original publication
- 2026
- Source retrieved
- 2026-09-20
What happened
Long Beach’s official CommsCoach privacy report card and PIA package document fully automated QA of ECC calls/radio/CAD, claim automatic evaluation of 100% of 911 calls, radio, and CAD records, process in AWS GovCloud with CJIS/SOC2 framing, delete unmapped GovCloud data within 24 hours, retain mapped workstation data for simulator/QA duration, and acknowledge OpenAI model bias risk while using binary yes/no evaluation criteria. Distinct from the April Hoodline feature already archived: this is the city’s governance artifact with retention and prohibited-use detail.
Why it matters
Gives SLED ECCs a concrete U.S. municipal template for post-call AI QA privacy controls—especially retention clocks and bans on unauthorized employee monitoring—when comparing vendors after CommsCoach entered the archive on 2026-09-18.
Evidence and measured results
Privacy report card retention section; PIA statements on 100% automated evaluations; GovCloud encryption; OpenAI bias link; prohibited uses including unauthorized employee monitoring.
Limitations and uncertainty
Package date not clearly stamped as a single ISO day on the report card face; performance benefit claims are purpose statements, not outcome trials; OpenAI dependency is a supply-chain risk.
Put this evidence to work
Lighthouse Advisory interpretation, grounded in this source. Enriched 2026-09-19; this does not change the original publication date. Each role follows the same checklist so takeaways compare across sources.
Sales
Role takeaway
- Customer problem
- ECCs want AI QA coverage but lack a public privacy/retention pattern that councils and unions will accept.
- Stakeholders
- ECC directors
- City privacy/IT
- Unions
- City attorney
- Council
- Discovery questions
- Is there a public PIA?
- Unmapped deletion SLA?
- GovCloud/CJIS?
- Coaching vs discipline policy?
- Value hypothesis
- If a city can show GovCloud + 24h unmapped deletion + coaching-only policy, QA AI procurement clears privacy gates faster.
- Potential engagement
- Privacy report-card and retention-control design workshop for ECC QA AI.
- Unsupported claims
- Clinical outcome improvement
- Bias-free OpenAI scoring
- Safe live 911 AI
- FTE elimination guarantees
Pre-sales engineering
Role takeaway
- Fit
- Post-call automated QA on recorder+CAD data in GovCloud with strict retention bifurcation.
- Architecture and integration
- Ingest recorder/CAD; map to workstation identities; score via binary rubric on LLM backend; delete unmapped blobs in 24h.
- Prerequisites
- GovCloud account
- Recorder/CAD feeds
- Identity mapping
- PIA/PUP
- Constraints
- OpenAI bias/supply chain
- Binary rubric may miss nuance
- Not for live answering
- Security
- Encryption at rest/in transit; access audit; CJIS/SOC2 controls; retention jobs monitored.
- Proof of value
- Deletion job logs
- Access audit samples
- Human re-score agreement
- Override rate
Delivery
Role takeaway
- Work
- Complete PIA, configure retention, train coaching use, publish report card, monitor deletion jobs.
- Dependencies
- Vendor GovWorx/CommsCoach
- AWS GovCloud
- City privacy office
- Union notice if required
- Ownership
- DPEC with Technology & Innovation privacy office oversight.
- Skills and adoption
- Supervisors use AI scores for coaching conversations, not silent discipline.
- Governance checkpoints
- PIA approval
- Quarterly retention audit
- Bias complaint pathway
- Acceptance criteria
- 24h unmapped deletion evidenced
- Public report card posted
- Written prohibited-use list enforced
- Risks
- Retention job failure
- Scope creep to live AI
- Union grievances over monitoring
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?
Logging-recorder + CAD export into GovCloud serverless QA; mapped vs unmapped data paths with different retention; OpenAI models behind binary rubric scoring.
Governance
Who approves, reviews and stays accountable for outcomes?
Public report card + PIA/PUP process; notify privacy office; define coaching vs discipline boundaries; document generative-AI guidance adherence.
Security and privacy
What data, permissions and controls need testing?
24-hour deletion of unmapped audio/data; GovCloud encryption at rest/in transit; CCPA/Delete Act references; residual model-bias risk disclosed via OpenAI article link.
Accessibility and workforce
Who is affected, and what skills or accommodations follow?
Binary scoring and simulator features can broaden QA coverage but may miss nuanced communication quality affecting LEP callers unless rubrics include accessibility items.
Procurement
What should contracts, pricing and exit terms secure?
Require GovCloud/CJIS, deletion SLAs, and prohibited-use clauses in ECC AI QA contracts.
Operating model
Which teams own the service once it runs?
Shifts QA from sparse manual sampling toward near-census review with human coaching interpretation.
Publication history
- 2026-09-19Emergency Services · Issue 145 resources
Stable resource ID: longbeach-commscoach-privacy-report-card-2026