Cortez City Council this week adopted an artificial intelligence policy applicable to city staff.
“This is all new territory for us,” said Councilor Bill Lewis, reviewing the eight-page document built off of policies from Boulder County; San Jose, California; Boston; Seattle; and Washington state.
The policy directs city employees to keep records of prompts used to generate material for their work. Permitted uses of AI include drafting internal memos or reports, brainstorming ideas and policy options, translation, correcting grammar and building customer service chatbots.
“Treat outputs as drafts; verify facts; screen for bias or harm,” reads the policy. “AI-generated content may never be relied upon as authoritative or final without human oversight and verification.”
AI assistance must be disclosed if it meaningfully contributes to publicly released content. City staff are responsible for verifying and editing the output.
“Please think of this as a living document as employees who use AI tools get a better understanding of them and the tools themselves evolve,” said Information Technology Director Jay Rohrer, who presented the draft policy.
The policy comes 3½ years after the public launch of ChatGPT. City spokeswoman Kelly Codner said in a statement that AI use has become more prevalent among city staff and there were no city standards for AI use prior.
Under the policy, staff are directed to use systems approved by the IT director or the city manager, a full list of which has yet to be determined.
“As employees gain a better understanding of how AI can assist them in their job, they will be exploring different tools that are available to them. Therefore, each list will naturally evolve over time,” Codner said.
Staff cannot input sensitive information, like health and criminal justice records into unapproved systems.
Employees are also restricted from generating videos or images representing a real person or city official or using AI to generate communications regarding legal, disciplinary and personnel decisions.
Law enforcement has an additional set of regulations stipulating that AI cannot be used for investigative decisions without formal approval and officials cannot input identifying information, such as social security numbers and criminal records, into public generative AI tools.
“This is a document that’s always going to be changing,” Lewis said. “I think the more usage we get from this, the more policies that are going to follow with that.”
avanderveen@the-journal.com
