The use of artificial intelligence in government operations is now restricted in Midway City.
City officials say the new ordinance will safeguard cybersecurity and ensure human approval of AI work.
Officials voted to restrict some uses of artificial intelligence, including the types of AI platforms allowed, for cybersecurity purposes. Instead, city staff can only use approved services with data encryption, role-based access controls, audit logging and administrative oversight.
City Attorney Corbin Gordon presented the ordinance, which the City Council previously discussed during a work meeting in March, on Tuesday.
The ordinance states that OpenAI platforms like ChatGPT are not permissible for use in city business. Gordon said in March that the city should be wary of OpenAI continually learning from vast amounts of data over time. He said it’s “not a great idea” to rely on the platform because of the data collection utilized by certain AI models.
AI is not allowed when confidential or privileged information is involved. Gordon said AI should not be making “the decisions for us.”
Staff can only use approved enterprise systems, and there must be express authorization for the use of external AI models. Instead, human approval and review are required and necessary for city business. Gordon said that guidance covers Midway legally if any issues come up.
Mayor Craig Simons said the restrictions are a matter of cybersecurity for the city.
Some uses of AI are allowed for basic administrative functions. That includes tasks like drafting preliminary meeting minutes, summarizing documents or reports, assisting in document organization and drafting non-final correspondence or internal documents.
Heber City Manager Matt Brower presented a similar AI policy to the Heber City Council in the spring. Wasatch County currently has no ordinance directing AI policies, and the topic was removed from a recent council agenda.