NEW YORK – The U.S. Department of Health and Human Services on Thursday announced it is supercharging its use of artificial intelligence to police how states and other recipients of federal health dollars are auditing their programs. The move is intended to tamp down risks of fraud and save the government money.
The department will use ChatGPT and other AI tools to analyze audit reports from all 50 states on an ongoing basis, said Gustav Chiarello, the assistant secretary for financial resources who is leading the new program.
“It’s classic big government: Everyone files an audit and it lands with a thud and no one does anything about it,” Chiarello said in an interview. “Here, with AI, we’re able to dig into it.”
The move builds on the department's embrace of generative AI for investigating state Medicaid programs, automating administrative tasks and editing text. AI tools can be a powerful aid in finding patterns or problems across large documents, but critics say the government should use them with caution because they frequently make mistakes and can have unintended biases.
The Trump administration and Vice President JD Vance’s anti-fraud task force have spent recent months promoting efforts to crack down on fraud in the Medicaid and Medicare programs as well as in student loan applications and other areas. Those efforts have also involved using AI technology to flag likely fraud, Federal Trade Commission Chairman Andrew Ferguson said recently on Fox News.
States, local governments, nonprofits and higher education institutions that spend at least $1 million in federal money a year are required to submit annual audits. The new initiative will use AI to analyze those audits from HHS-funded programs, including state Medicaid programs and federal grantees in research, addiction services and more, Chiarello said.
Recipients that do not file the required reports or resolve problems in them could face a loss of money.
Critics have blasted the administration's anti-fraud efforts, noting most have been targeted at Democratic states and at times have reflected a tendency to attack first and gather the facts later. On at least one occasion, the administration acknowledged to The Associated Press that it made a major mistake in data it had used to help justify a New York Medicaid fraud investigation.
Asked about safeguards against the AI tools making mistakes, Chiarello noted that officials were evaluating public reports rather than uncovering new information. He said the tools were intended to make grantees better stewards of federal dollars.
The department said it has sent letters to governors and treasurers in all 50 states alerting them to the new initiative. The program was first reported by The Wall Street Journal.
Chiarello said he has been in touch with his counterparts in other federal departments in hopes that they follow his lead.
“It would be fairly easy for the other agencies to use our technology and jump on it,” he said.