This article builds upon our previous reporting with the Government Technology AI Tracker, providing the latest developments in AI policy and legislation across the U.S. The tracker is updated with the latest information from our ongoing coverage, legislative action and updates from state government websites.
Many are transparent about what they’re working on. North Carolina has created an AI Corner with information about how AI is being used in the state. Utah has launched an AI Learning Lab for its new Office of Artificial Intelligence Policy. The California Department of Technology has published a new Artificial Intelligence Community web page.
The following are noteworthy trends spotted within our Government Technology tracking data.
BEYOND TASK FORCES: NEW PATHS TO AI POLICY
By the end of 2023, a few states made headlines for launching or starting the process of setting up AI councils or task forces for state governments.
A year later, at least 33 states have some kind of AI council, task force or working group. The region that has the most of these groups is the South, specifically the South Atlantic division.
But the absence of a task force doesn’t necessarily mean there’s not a lot of AI planning happening in a state. In some cases, the absence of a task force is the result of state-specific or legislative challenges.
The Alaska Senate State Affairs Committee introduced Senate Bill No. 262, proposing the establishment of a state AI task force within the Department of Commerce, Community and Economic Development. However, the bill died in committee.
Meanwhile, a few states have managed to publish a state policy without an established AI task force, such as North Carolina. Meanwhile, Utah’s legislation called for the creation of an AI office rather than a task force to guide the creation of a state policy.
With or without a task force, states in about half of the country have created some kind of AI state-specific framework or guidance. However, states with a task force are more likely to have created a policy, as 19 of the 25 states with a policy or draft have task forces.
As the prevalence of written and published policies has grown, so have restrictions and guardrails on how AI can be used in some states. There are at least 10 states with a publicly viewable policy that stipulates some kind of restrictions or requirements that must be met before using or launching an AI program. Typically, the policy requires use of AI to be reviewed by a specific council or department and approved before use.
AI INVENTORIES AND TRAINING PROGRAMS EMERGE
At least 13 states, primarily located in the South, have written into their state AI policy, or have been legislatively required, to keep an inventory of all AI systems that the state uses.
The required information for these lists of AI tools varies across states. Typically state IT departments are tasked with conducting audits, while each agency is responsible for reporting correct information. Some states require the inventory to be publicly posted or included in an annual report.
One emerging trend is states offering some type of AI training, or training guidance by compiling several sources into one list. There are now eight states with some kind of training, many geared for different types of audiences.
Oklahoma partnered with Google to offer a course for residents, while Maryland is working with InnovateUS to provide AI education for state employees.
UNCHARTED WATERS: A SPLIT APPROACH TO LEADERSHIP
One topic where a definite trend hasn’t appeared: who is in charge of a state’s AI.
While some states have created a leadership position to guide AI implementation, others are creating entire offices. New York, New Jersey and Georgia have named individuals as AI leaders in the state, while Utah, Connecticut, Rhode Island and Vermont have created leadership offices.
The leadership challenge came up at the National Association of State Chief Information Officers (NASCIO) 2024 Annual Conference, with President Jim Weaver suggesting a leadership office rather than a singular position might be a more efficient and less bureaucratic option.
