How is LFUCG using artificial intelligence? Council will review city's AI policy this week

This week, Lexington's Urban County Council will hear a presentation covering the local government's artificial intelligence policy, which outlines acceptable and prohibited uses for the technology, along with other uses it's currently considering.

The presentation from Chief Information Officer Liz Rodgers is on the agenda for Council's General Government and Planning Committee, which meets Tuesday, June 2nd at 1 p.m. The meeting will take place in the Council Chamber of Lexington's Government Center. You can attend in-person or watch it online via LexTV.

As artificial intelligence becomes more widespread, the Lexington-Fayette Urban County Government is exploring uses of the technology in its work.

In early May, Councilmembers heard from LFUCG Code Enforcement Director Oliver Steele, who described how the department wants to use artificial intelligence to identify enforcement "hotspots" around town and target them more proactively.

LFUCG's existing artificial intelligence policy dates to October 2025, and is subject to change as the technology evolves. The policy lays out some general ground rules for AI use. It outlines what is and isn't allowed, establishes requirements for human oversight, and bans inputting confidential, regulated, or proprietary information into any AI tool.

The policy's scope is broad, applying to LFUCG employees, interns, consultants, contractors, and third-party vendors. It covers AI usage "on any device, network, or system where the activity involves LFUCG data, systems, or business processes."

How is LFUCG using AI? What is it considering?

The presentation notes there are several acceptable uses for the technology, provided certain accountability and transparency conditions are met. They include:

• Drafting and content support (writing assistants)
• Process and workflow efficiency (organizing, formatting, and summarizing materials)
• Data analysis and research support (extracting patterns, assisting with research)
• Employee training and enablement (learning experiences and simulations)

The presentation also notes several AI use cases currently under review, including AI-assisted review of planning and permitting documents, live language interpretation for voice calls, and AI agents that can assist with resident requests.

Prohibited AI uses under LFUCG's policy

The following uses are strictly prohibited under the policy:

• Use of public or unapproved AI tools. AI tools have to be approved for use by LFUCG's Department of Information Technology. The policy states the: "Use of publicly accessible AI tools, whether free, browser-based, trial version, or via subscription are not permitted for any work-related activity, regardless of data sensitivity or perceived risk.". Doing so violates privacy and security standards.
• Inputting or handling sensitive data. Users aren't allowed to input sensitive data or customer-specific data into any AI tool, regardless of the purpose.
• Autonomous actions or automation without oversight. Agentic AI – or artificial intelligence that can act with little to no human involvement – cannot be used to make decisions, approve transactions, execute scripts or perform tasks without direct human initiation, oversight, and review.
• Impersonation and misrepresentation. Using AI tools to mimic or impersonate LFUCG employees, customers, vendors or regulators in any format is prohibited. Examples include creating or distributing deepfakes, fake voices, avatars or images.
• Unsafe, misleading or unsanctioned outputs. There's a broad rule against using AI tools to "generate content that is inaccurate, incomplete, biased, deceptive, defamatory, discriminatory, or in violation of legal, regulatory, or ethical standards."
• Bypassing internal controls. Uses cannot use their personal devices, incognito browsers, proxies, virtual private networks (VPNs) or other tools to get around LFUCG policies or restrictions on AI use. That includes attempting to enable experimental features, plug-ins or agentic behavior in AI tools without approval.

You can read the full policy and review the presentation to Council starting on page 3 of the meeting packet.