The Department of Government Efficiency overly relied on artificial intelligence to terminate contracts at the Department of Veterans Affairs, top Democrats on the House and Senate VA panels said on Tuesday.
During a “spotlight forum” led by Sen. Richard Blumenthal, D-Conn. — the ranking member of the Senate Veterans’ Affairs Committee — lawmakers pressed their concerns that VA has arbitrarily cancelled hundreds of contracts in recent months without the necessary oversight needed to determine their importance to the agency. The forum did not include any Republican lawmakers.
The DOGE cost-cutting unit has been tasked by President Donald Trump with slashing spending across the federal government, which has included mass layoffs across some agencies and widespread contract cancellations. In line with these efforts, VA Secretary Doug Collins announced that the agency is considering laying off as many as 83,000 employees across its operations this year to return to its 2019 staffing levels. Collins has stressed that the efforts are designed to streamline VA’s operations and would not affect mission-critical programs.
As part of its cuts, VA also announced in March that it was terminating “585 non-mission-critical or duplicative contracts,” which it said were identified “through a deliberative, multi-level review that involved the career subject-matter expert employees responsible for the contracts as well as VA senior leaders and contracting officials.” The agency also said the news represented “the first step” in a review of VA’s roughly 90,000 contracts.
Democrat lawmakers have expressed alarm about the scale of terminated contracts, alleging that VA has intentionally slow-walked the full disclosure of the cancelled agreements in response to their requests for information. VA, for its part, said it would not cancel any contracts that impact veteran services and that it is merely targeting those that are wasteful.
The lawmakers’ concerns about how and why specific contracts were identified for termination, however, were amplified by a June 6 ProPublica investigation that said a software engineer with DOGE built an error-prone AI tool that was used to identify contracts for “munching” — or cancellation — by VA. The DOGE employee, who has since been let go after he publicly spoke about finding less inefficiencies in government than he expected, told ProPublica that “mistakes were made” with his AI tool.
Blumenthal referenced the news report during Tuesday’s forum and said that, while he was in favor of cutting waste at VA, “the way to focus on it is thoughtfully and factually, not to use AI and identify keywords in the contract that then result in wholesale slashing.”
The former DOGE employee said analyzing VA’s almost 90,000 contracts within the 30-day period outlined by President Donald Trump’s February executive order that mandated a review of agencies’ “spending on contracts, grants and loans” would have been impossible without developing the AI tool.
Rather than carefully evaluating which of the cancelled VA contracts were duplicative or unnecessary, Blumenthal said this approach relied on feeding contract information into the AI tool for analysis without the necessary level of oversight. The contract cancellations, he claimed, were “applied arbitrarily without that individualized evaluation based on human beings looking contractors and veterans and VA workers in the eye and determining where cuts make sense.”
One of the witnesses at Tuesday’s forum, Benjamin Ambrose, was part of the team at Aptive Resources that had a contract to support VA’s efforts to modernize its electronic health record system. The new EHR system has been beset by a host of challenges since it was first deployed in 2020, but officials are hoping to speed up deployments of the software starting next year. Despite VA’s ongoing commitment to the EHR modernization effort, Ambrose said the company’s contract with the agency was cancelled by DOGE.
Rep. Mark Takano, D-Calif., the ranking member of the House Veterans’ Affairs Committee, said it seemed as though “the AI algorithm looked for some keywords in this contract [and] decided that this contract should be terminated,” noting its connection to the broader EHR modernization project and the possibility that it was identified as a redundancy.
He added that “any sane efficiency effort” would have assumed the AI tool “was going to identify some similarities, but they didn’t do their homework.”
“They were more interested, in my view, of getting some sort of headline, some sort of big number of the contracts [cut],” Takano added. “They wanted to run up the numbers instead of really thinking about who was getting hurt.”
