AI Reliability Concerns Lead Grand Traverse County to Propose Formal Oversight
GRAND TRAVERSE COUNTY — The Grand Traverse County Information Technology Department is proposing the establishment of a Center of Excellence for Artificial Intelligence to oversee the secure adoption of AI technologies across the local government. Information Technology Director Cliff DuPuy will present the proposal to the Board of Commissioners on Feb. 11, 2026, requesting a $118,000 investment for the 2027 fiscal year.
This proposal arises in response to a recent surge in federal data security breaches and growing concerns regarding AI reliability. County officials cited the theft of billions of records from the National Public Data and hundreds of court cases involving AI-generated misinformation as primary reasons for creating a formal governance framework. The center would provide security oversight and documented processes for the county’s 600 employees and 95,000 residents.
Budget Breakdown
The proposed $118,000 budget for the 2027 fiscal year includes:
- $75,000 for one full-time staff position
- $18,000 for technology licenses
- Remaining funds allocated for training, consulting services, and new security tools designed to detect AI hallucinations and monitor prompt injections
IT officials expect this investment to achieve a positive return within 18 to 24 months through efficiency gains and cost avoidance.
The AI Reliability Crisis
The proposal highlights an “AI Reliability Crisis”, noting that 486 court cases worldwide have involved AI hallucinations, with 324 of those occurring in United States courts. According to a 2024 Stanford study cited in the proposal, organizations with documented AI governance achieve a 96% reduction in AI errors. Without these processes, the IT department warned the county would have no defense in litigation or protection from insider threats.
DuPuy argued that the county must formalize its AI oversight to protect public trust. “The question is not whether we need AI governance—recent events answer that definitively,” DuPuy stated. “The question is whether we act now, while we’re ahead, or wait until we’re explaining to taxpayers why we didn’t.”
Updated Acceptable Use Policy
The initiative includes an updated Acceptable Use Policy that mandates “human-in-the-loop” validation for all AI-generated content. Under these rules, county employees are prohibited from uploading personally identifiable information, health records, or law enforcement data to public AI platforms. The policy also requires users to disclose when AI has assisted in the preparation of official documents or reports.
Four Pillars of the Proposal
The county’s plan is based on four pillars: governance, expertise, innovation, and security. This structure is intended to protect sensitive data from 911 dispatch, courts, and health departments. The proposal cites success stories from other regions, including Williamson County, Texas, which saved $900,000 and 36,000 staff hours by using AI for property assessments.
Justification for Immediate Action
Recent federal security failures were used to justify the immediate need for the center. These include a December 2024 breach at the Treasury Department that exposed 3,000 files and the theft of 2.9 billion records from National Public Data. The IT department noted that even unclassified systems can face catastrophic breaches if proper security protocols are bypassed.
If the Board of Commissioners approves the proposal, the first steering committee meeting will be held within 30 days. The implementation roadmap scheduled for 2027 includes a foundation phase to assess current AI usage and a capability-building phase to launch pilot projects with hallucination detection.
