A Fulton County commissioner alleged Tuesday that four ordinances on Wednesday's meeting agenda were drafted with artificial intelligence, setting off a public dispute with a fellow commissioner over the responsible use of the technology in lawmaking. The clash raises immediate questions about transparency and vetting standards for legislation that could affect residents across Georgia's most populous county.
Commissioner Dana Barrett issued a press release claiming that three ordinances sponsored by Commissioner Marvin Arrington Jr. registered as 100% AI-generated when she submitted the text to GPTZero, an AI-detection platform. A fourth ordinance received an 89% score. "I didn't need a detector to tell me that," Barrett said. "I saw it with my own eyes. The formats were different. The word choices were different. These are not normal ways our resolutions are written."
Arrington acknowledged using AI to assist with drafting but denied the ordinances were written entirely by the technology. He described the process as a hybrid of AI-assisted drafting and human legal and policy editing. Barrett's goal, she said, is not to ban AI but to ensure it is used responsibly. "What's happening right now is dangerous," she said. "Shoving things through on the agenda that could get passed, because somebody has four votes, without thoughtful consideration or a thoughtful read is dangerous."
Detection tools yield conflicting results
GPTZero's output is not definitive. When the same proposals were analyzed using ChatGPT, the response characterized Arrington's ordinances as "likely a hybrid of AI-assisted drafting and human legal/policy editing." GPTZero itself acknowledges that no AI detector is completely accurate and that its results should not be used as a final verdict. Barrett's reliance on a single detection score highlights the gap between available tools and the certainty policymakers and the public might expect from them.
The dispute echoes broader challenges in AI for Legal applications, where authorship verification remains an unsettled question. Detection platforms can produce differing results from identical text, and their probability scores do not reveal who conceived, directed, or approved a policy.
Arrington defends the practice, plans AI governance policy
Arrington issued a lengthy statement pushing back against Barrett's characterization. "Commissioner Barrett used artificial intelligence to accuse others of using artificial intelligence. Imagine that. If that is not the pot calling the kettle black, I do not know what is," he wrote. Arrington, a practicing attorney of 30 years and a 12-year county commissioner, argued that AI is a tool comparable to legislative templates, search engines, and legal research platforms.
The ordinances address food security, public comment standards, and ethics board reform. Arrington said he developed the ethics resolution partly from personal experience with a complaint he said was wrongfully held over him for four years before a court ordered it dismissed. He plans to introduce a comprehensive policy governing Fulton County's responsible use of AI at the next board meeting and has already circulated the proposal to county staff for input. For policymakers navigating these questions, AI Learning Path for Policy Makers resources can help establish guardrails before legislation reaches a vote.
No current restrictions on AI use in legislation
Fulton County has no rules restricting commissioners from using AI to draft legislation. Barrett wants that to change. The four ordinances were scheduled to go before commissioners Wednesday, where the debate was expected to shift from authorship to the substance of the proposals themselves. Barrett said any objection should identify specific provisions that are legally defective or harmful, rather than relying on a detection score to dismiss the legislation.
Why this matters for legal professionals
For attorneys and legal staff who draft, review, or challenge local ordinances, the Fulton County dispute surfaces a practical risk: legislation shaped substantially by AI without clear disclosure or human validation can introduce ambiguities, drafting errors, or policy gaps that courts later have to untangle. The conflict also signals that AI governance policies are moving from abstract discussion to immediate necessity inside government bodies. Legal professionals advising public agencies should anticipate questions about authorship, detection reliability, and the standard of care expected when elected officials use generative tools to produce binding law.
Your membership also unlocks:
