(TNS) -- Citing security and privacy concerns, the Michigan Senate is now blocking lawmakers and staffers from using Senate-issued devices to access some artificial intelligence tools, including the popular chatbot ChatGPT.
The Michigan Senate Information Services announced the internal policy on Monday in an email, as government agencies across the country weigh how to interact with new AI technology. The message from the Michigan Senate mentioned the possibilities of "draft legislation" and constituent information being entered into AI platforms.
ChatGPT, which launched in 2022, can create images and write articles and text for emails in response to prompts provided by users and can analyze data.
"While (Senate Information Services) understands that AI tools are useful in many applications, including the work being conducted by Senate offices, these tools pose security and privacy risks that could have significant implications for our organization," the Monday message from the Michigan Senate said.
As of Friday, Michigan senators, employees and interns who attempt to access AI-focused tools, including Chat GPT, Gemini, Claude and Perplexity, on Senate-issued devices "will encounter an error message and will not be able to access the resource," according to the email from the Senate.
The "cybersecurity update" said many AI resources store data submitted by users, meaning the information could be accidentally shared. Plus, the message warned about the submission of information, including "draft legislation" that wasn't ready to be released publicly.
"The entry of proprietary and/or sensitive information into an AI tool, including constituents' personally identifiable information, draft legislation, policy research or other material not intended or ready for public consumption, risks exposure to other users of the AI tool or incorporation into the AI's training data," the Senate email said.
Todd Cook, director of the Senate Business Office, didn't immediately respond Tuesday to a question about whether any specific incidents prompted the restrictions on AI tools.
As of Tuesday, it appeared the Michigan House, which has its own business and technology policies, hadn't put in place limitations on ChatGPT or similar technology because multiple state House staffers told The Detroit News they could still access ChatGPT.
Democrats currently hold majorities in both the Michigan House and Senate. However, Republicans will take control of the state House in January.
In March, the White House Office of Management and Budget directed federal agencies to establish AI governance boards and to implement safeguards for using AI by Dec. 1.
ChatGPT's website says its "users can control whether their content is used to train our AI."
"When you share your content with us, it helps our models become more accurate, safer and helps to solve your specific problems," the website adds.
The Michigan Senate email said the Senate wasn't restricting access to most websites or tools that include only "limited AI capabilities" and would make Microsoft 365 Copilot, an AI chatbot developed by Microsoft, available by March 2025.
©2024 The Detroit News, Distributed by Tribune Content Agency, LLC.