When a city or county puts AI on a decision about a resident, the explanation and the liability stay local. AVAAS certifies the system.
State agencies, counties, and city governments now run AI on benefits eligibility, licensing and permits, housing and voucher decisions, code enforcement, and resident-facing chatbots. There is no federal floor for most of it. When the system is wrong, the resident is harmed and the jurisdiction carries the liability. AVAAS certifies the system at the point it decides for a person.
The decision point in local government
In state and local government, the AI acts when it decides eligibility for a benefit, approves or denies a permit or license, scores a tenant or a voucher applicant, flags a property or household for enforcement, or answers a resident’s question about the law on an official site. Each of these is a decision a person can be harmed by and a court can review.
A city chatbot that gives unlawful guidance, or an eligibility tool that denies a benefit by mistake, still speaks with the authority of the government that deployed it.
What keeps state and local agencies exposed
A denial the agency cannot explain
Procedural due process reaches benefit, license, and housing decisions whether a person or a model makes them. An unexplained automated denial is both a legal liability and a human harm. Michigan’s MiDAS system wrongly accused tens of thousands of residents of unemployment fraud at an error rate above 85 percent when it adjudicated by algorithm alone.
State and local law is what applies
OMB M-25-21 binds federal agencies, not cities and counties. State, county, and municipal AI sits under state law, anti-discrimination and fair-housing rules, and constitutional due process, and the new state automated-decision regimes such as California ADMT and the Colorado AI Act begin to bite in 2027.
Official answers that break the law
New York City’s MyCity chatbot, on the city’s own site, told business owners they could pocket workers’ tips and refuse tenants with housing vouchers, both unlawful, and the city left it running long after the problem was exposed. A tool the agency cannot inspect can hand residents illegal guidance, and relying on a vendor does not move the liability.
Evidence the system behaves the way the public can hold you to
Can an automated denial be explained to the resident and a court?
AVAAS evaluates whether the system produces specific, accurate reasons a person and a reviewer can actually follow.
Does a resident-facing system give lawful, accurate answers?
AVAAS evaluates how the system behaves at the point its output reaches a resident, including where it can confidently state something untrue or unlawful.
Does the system produce disparate outcomes?
Five structurally independent validators test for demographic disparity and failure patterns using causal attribution.
You get documented, third-party evidence that an eligibility, licensing, housing, or resident-facing system behaves the way the public sector is accountable for, without a federal standard to lean on.
Related AVAAS coverage: California ADMT · Colorado AI Act · GDPR Article 22. Or run the free Regulation Checker to see what applies to you.
See where your local-government AI creates liability.
Tell us where automated systems decide eligibility, permits, housing, enforcement, or answer residents, and we will scope an AVAAS certification to the exposure.
Ready to start now? Certify Your AI → or email [email protected]