The Palm Desert City Council is scheduled to vote Thursday, October 8, on a citywide artificial intelligence policy for city work. The regular session starts at 4 p.m. in the Council Chamber at 73-510 Fred Waring Drive. The policy is item 13.f on the consent calendar, which the agenda describes as routine matters that may be approved by one motion; the council can remove an item for separate discussion. If the council adopts the resolution, the policy takes effect immediately.
Staff say the city has relied on administratively approved guidance written for generative AI tools, and that AI has since expanded well beyond what that guidance covers. The city created an Artificial Intelligence Officer position in March 2026, and the staff report, prepared by officer Joe Balich, calls the governance policy that position’s first major deliverable.
The policy sorts AI uses into three risk levels by what the use does and the boundary it operates in, not by the tool alone. Low-risk uses, such as drafting, research, summarization, translation and analysis with a person reviewing the output, need no additional approval when they stay within an approved boundary. Moderate-risk uses need the responsible department director’s approval. They include using sensitive or restricted city information or inferring sensitive personal data, communicating directly with the public without a person approving each message, and establishing or materially expanding a connection that gives AI write access to a city system or authority to act. High-risk uses need the director’s approval and the city manager’s written approval before they begin. They include AI used to recommend or materially influence a decision that materially affects a person’s rights, benefits, services, permits, enforcement or employment; AI that communicates or acts without prior human approval in a way that materially affects a person; and autonomous action that creates a legal or financial commitment for the city or poses a significant legal, financial, security or operational risk. When more than one level applies, the higher one governs.
For all city work, the policy requires approved AI tools used through city-provided accounts and within approved boundaries. The person responsible for the work must review and verify AI-assisted output before relying on it, and a person must make any decision that requires judgment or discretion. The city may not use AI to create generalized social scores of people. No one may use AI to falsify a city record, falsely attribute a statement to a person or the city, or get around security controls or required approvals. Using sensitive city information to train or modify an AI model requires the city manager’s written authorization.
Three commitments face the public. The city would publish and maintain a public inventory of approved AI tools and their general purposes. AI that communicates directly with the public for the city must identify itself as AI, give a practical way to reach city staff and meet accessibility requirements. And when a high-risk use materially influences a decision about an identifiable person, the city must tell the person AI played a material role and give a practical way to seek human review or correction.
The policy also addresses records. It says AI prompts and outputs used solely as working material are preliminary drafts or notes the city does not retain in the ordinary course of business. Before disposing of one, the person responsible must make sure the substance of any information relied on is in an appropriate city record kept under the city’s retention schedule. A prompt or output that is itself the city’s record of a communication or action must be retained under that schedule, and nothing may be disposed of while it is under a legal hold or a pending records request, or when preservation is otherwise required, including for pending or reasonably anticipated litigation.
The policy covers employees, interns, volunteers, elected and appointed officials and others authorized to use city-provided AI accounts, when their use involves city work or sensitive or restricted city information. Contractors are governed by their agreements, and personal or campaign use of AI with personal resources and public information is outside the policy.
Staff say adoption has no direct fiscal impact to the General Fund and does not authorize buying any AI product or service; any future purchase would follow the city’s normal budget and procurement processes. The city manager would approve implementation guidelines and set schedules or transitional measures for existing AI uses, and the AI officer would review the policy and the program at least annually.
The packet does not say how the council will vote. It also does not say which tools are approved, what the approved boundaries are, or when the public inventory would be published. The policy leaves boundaries to implementation guidelines the city manager would approve, and the resolution number and adoption date are blank.