Responsible AI in Local Government: From Principles to Action
Generative AI is showing up across city and county workflows, from constituent questions to core operations. A Center for Democracy & Technology (CDT) panel laid out a clear path: focus on transparency, reliability, governance, privacy, and compliance - and build systems people can trust.
Regional Government Services Authority Deputy Executive Director Rich Oppenheim put it plainly: human-centered AI has to match how government actually works. As he said, "Responsible AI is about building systems, including training, establishing decision rights, accountability and transparency - that assume imperfect humans working with imperfect AI, operating under real conditions, but still working to produce outcomes the public can trust."
The five priorities governments agree on
- Public transparency and stakeholder engagement
- Accuracy and reliability
- Governance and coordination
- Privacy and security
- Legal compliance
These aren't slogans. They're guardrails that keep pilots from turning into problems - and help staff ship value without creating new risks.
Learn more about CDT's work on AI policy at cdt.org.
Human oversight isn't optional
AI can draft, summarize, predict, and route. It can also hallucinate, drift, or break on edge cases. You need human review, clear decision rights, and an escalation path before anything touches residents, benefits, or public records.
Start with low-risk assistive use (drafts, search, summarization). Add structured review. Measure outcomes. Only then consider decisions that affect services or eligibility.
Rural access and uneven capacity
Many rural communities still lack broadband or dedicated IT teams. That gap limits who benefits first. The takeaway from the panel: don't wait for perfect capacity. Start small, prove value, and grow your policy and practice as you go.
Pick use cases that work offline or with low bandwidth, and share resources across jurisdictions when possible.
What it looks like in practice
- Transparency by default: Lebanon, New Hampshire maintains an online AI registry so residents can see what tools the city uses and why. Simple, public, and accountable.
- Train before you scale: Calvert County, Maryland produced internal videos on generative AI to set expectations and common rules before broader adoption.
- Fix the foundations: Alameda County, California uses AI to update legacy code - accelerating upgrades for systems that are essential to daily operations.
As National League of Cities' Christopher Jordan noted, vendors will promise everything. The real question: what problem are you solving, and how will residents feel the difference?
Practical next steps for city and county leaders
- Pick 3 use cases that save staff hours or improve response time (draft responses, meeting notes, document search).
- Set decision rights (who can approve tools, who reviews outputs, who owns incidents).
- Publish an AI registry with purpose, data sources, risk level, and contacts.
- Create a lightweight policy covering acceptable use, prohibited data, privacy, and record retention.
- Run small pilots with success metrics (quality, time saved, errors caught) and a stop/go decision.
- Procure with intent: require security, audit logs, model change notices, and exit clauses.
- Upskill staff with short trainings and prompts that fit each role.
- Mind rural constraints: favor tools that work with low bandwidth and shared services; seek state partnerships.
- Keep humans in the loop for anything that affects eligibility, compliance, safety, or rights.
About the discussion
The CDT webinar, "How Can State and Local Governments Responsibly Use and Govern AI in 2026," was moderated by CDT Policy Analyst Maddy Dwyer and featured Regional Government Services Authority Deputy Executive Director Rich Oppenheim, National League of Cities Senior Specialist on Urban Innovation Christopher Jordan, and National Association of Counties Telecommunications and Technology Legislative Director Seamus Dowdall.
Helpful resource for staff training
If you're planning internal AI upskilling, explore role-based programs that align with government workflows: Complete AI Training - Courses by Job.
Your membership also unlocks:
