DOGE’s Flawed AI Tool Misused VA Contracts
▼ Summary
– The Trump administration enlisted a software engineer with no health care or government experience to develop an AI tool for identifying non-essential VA contracts, labeling them “MUNCHABLE.”
– The AI tool used outdated models and made significant errors, such as inflating contract values—mistaking $35,000 contracts for $34 million.
– Over 2,000 contracts were flagged for cancellation, but the administration’s decisions remain unclear, with VA officials stating nearly 600 contracts have been canceled overall.
– Among the canceled contracts were critical services like maintaining a gene sequencing device for cancer research and blood sample analysis for VA projects.
– AI and procurement experts reviewed the tool and deemed it flawed, criticizing the use of AI for VA budget cuts as “deeply problematic.”
When the Trump administration sought to streamline operations at the Department of Veterans Affairs, they relied on an unconventional approach, a hastily developed AI tool created by a software engineer with no background in healthcare or government. This system, built under the Department of Government Efficiency (DOGE), was designed to flag non-essential contracts for cancellation, labeling them as “MUNCHABLE.” However, the tool’s methodology raised serious concerns among experts.
The AI relied on outdated and low-cost models, leading to significant errors in its assessments. Contract values were frequently misrepresented, with some agreements inflated to $34 million when their actual worth was as low as $35,000. Despite these flaws, the system identified over 2,000 contracts for potential termination. The VA has since canceled nearly 600 contracts, though the exact criteria for these decisions remain unclear.
Congressional Democrats have pressed for transparency, but details on which contracts were cut, and why, have been scarce. Independent review revealed that critical services were among those eliminated, including a contract for maintaining gene sequencing equipment used in cancer research and another supporting blood sample analysis for VA studies. Additional cuts affected programs aimed at improving nursing care standards.
After obtaining the AI’s code and flagged contracts, ProPublica consulted multiple AI and procurement specialists. Every expert agreed the system was fundamentally flawed, with several questioning the wisdom of using such technology to guide budgetary decisions at an agency responsible for veteran care. One described the approach as “deeply problematic,” highlighting the risks of automating complex policy choices without proper oversight.
The episode underscores broader concerns about the misuse of AI in government decision-making, particularly when applied to sensitive areas like healthcare and veteran services. Without rigorous validation and domain expertise, such tools risk causing unintended harm—raising questions about accountability in an era of increasing automation.
(Source: Ars Technica)
