More than 50 Maryland state agencies now use artificial intelligence in their operations, as the administration of Gov. Wes Moore accelerates adoption. The scope of deployment - spanning human services, transportation, health and the governor's own office - has sparked renewed debate over transparency, oversight and how AI shapes interactions between government and residents.
The state's public inventory shows agencies using both custom-built tools and commercial products such as ChatGPT, Claude, Google's Gemini and Microsoft's Copilot. The governor's office said it uses AI for drafting documents, analyzing data, handling public inquiries and reviewing records. Moore himself recently said he turned to Claude, Anthropic's AI chatbot, to answer budget questions and compare his administration's spending reductions with those of past governors.
Last month, Maryland launched an AI Innovation Lab to help agencies test new applications. Researchers describe the state as an aggressive government adopter of AI. But as the technology moves deeper into public services, they caution that residents often have no way to know when AI is involved, whether it influences decisions, or what recourse exists if something goes wrong.
Questions of transparency and human judgment
Doug Lombardi, a researcher at the University of Maryland Artificial Intelligence Interdisciplinary Institute, pointed to the tension between efficiency and accountability. "AI can be a really helpful tool, but the big concern is that it takes the human out of the equation of judgment and decision-making," he said. "Marylanders are going to be using AI whether they realize it or not."
Lombardi noted that the state's inventory identifies which tools agencies use but offers little detail about costs, benefits or downstream effects. He argued agencies should clearly explain "how you use it, when you use it, why you use it."
Henry Farrell, who studies artificial intelligence, data and democracy at Johns Hopkins University, said AI can help residents navigate complex programs. But he warned that relying on automated systems to sort data, generate outputs or assist with analysis can make government decision-making less transparent - and harder for the public to trace.
Guardrails, oversight and courtroom fallout
The governor's office referred questions about oversight to the Maryland Department of Information Technology. Nathan Miller, a department spokesperson, said any AI-generated content must be reviewed and approved by a state employee before it is used. "The State of Maryland is using AI as a support tool," he said in an email. "Responsibility for any official work product, public statement, or budget decision rests with the responsible state officials and employees."
Miller added that the state's Responsible AI Policy provides additional safeguards and requires extra review for high-risk AI systems. Last week, Moore named Michael Boyce, a former Department of Homeland Security and White House technology official, as senior adviser for responsible AI to help coordinate strategy across agencies. The appointment reflects a governance focus that echoes principles covered in the AI Learning Path for Policy Makers.
The consequences of insufficient disclosure have already surfaced in Maryland courts. Last year, the Maryland Appellate Court overturned a robbery conviction after prosecutors failed to properly disclose their use of facial recognition technology to identify a suspect. The court ruled the delayed disclosure prevented the defendant from adequately challenging the reliability of the AI-assisted evidence.
Lawmakers weigh the pace of change
Prince George's County Sen. Ron Watson, a Democrat on the Senate Education, Energy and the Environment Committee, said lawmakers have held regular discussions about agency AI use and the guardrails needed. Those conversations center on preventing bias, protecting personal information and ensuring fairness when AI appears in government processes.
Yet Watson cautioned that technology is moving faster than the legislative process. "The implementation is the thing that we really, really have to worry about," he said, adding that AI oversight cannot be solved by the Maryland General Assembly alone. "This is a job for industry, this is a job for academia."
As Maryland deepens its use of AI across government - part of a broader AI for Government trend - researchers say the central test won't be whether the technology improves efficiency. It will be whether residents can still understand, challenge and appeal decisions that affect their lives.
Why this matters for government professionals
Maryland's experience shows that AI deployment at scale demands clear governance, documented accountability and a deliberate effort to keep decision-making transparent. For public sector professionals, the message is direct: adoption is accelerating, but the burden of proof sits with the agency - not the resident - to demonstrate that automated tools are used fairly, reviewed by humans, and subject to the same avenues of appeal as any other government action.
Your membership also unlocks:
