Federal agencies would need to disclose when they use artificial intelligence systems to make “critical decisions” and inform people when they interact with an AI employed by the government under a bipartisan bill unveiled by a group of senators Thursday.
The legislation, shared first with POLITICO, would also direct agencies to create an appeals process that would require human review of decisions made by AI systems that may have harmed individuals. The bill, known as the Transparent Automated Governance Act, lays out a range of automated decisions that would be subject to disclosure and review — including those related to federal benefits, employment, essential utilities, housing, education, immigration and asylum.
Sen. Gary Peters (D-Mich.), chair of the Homeland Security and Governmental Affairs Committee, introduced the bill with Sens. James Lankford (R-Okla.) and Mike Braun (R-Ind.).
In a statement, Peters said that AI “is already transforming how federal agencies are serving the public.” The senator said his bill “will ensure taxpayers know when they are interacting with certain federal AI systems and establishes a process for people to get answers about why these systems are making certain decisions.”
A Peters aide said the committee plans to mark up the legislation next week.
The TAG Act comes several weeks after a Governmental Affairs Committee hearing that examined whether Congress should pass legislation to rein in the federal government’s use of AI. At the time, several witnesses suggested lawmakers should consider requiring more transparency on the federal government’s use of AI.
Last month Sen. Rand Paul (R-Ky.), the committee’s ranking member, told POLITICO he would struggle to work with Peters on AI legislation until Democratic lawmakers agree to address the potential for AI systems to enable online censorship. But other GOP senators, both on and off the committee, don’t appear to have a similar hang-up.
“No American should have to wonder if they are talking to an actual person or artificial intelligence when interacting with the government,” said Braun, who introduced an AI workforce bill with Peters that advanced out of the committee last month . Braun, who is not a member of the Governmental Affairs Committee, added that agencies need to be “proactive and transparent with AI utilization and ensure that decisions aren’t being made without humans in the driver’s seat.”