DOGE developed a flawed AI tool called 'munchable' to identify non-essential Veterans Affairs contracts for cancellation, which produced significant errors including contract value hallucinations and led to the cancellation of contracts supporting veterans' medical research and care.
In early 2023, the Department of Government Efficiency (DOGE) enlisted software engineer Sahil Lavingia to build an AI tool to identify non-essential contracts at the Department of Veterans Affairs for potential cancellation. Lavingia, who had no healthcare or government experience, developed the 'munchable' tool in just two days using outdated and inexpensive AI models. The system analyzed only the first 2,500 words of contracts and produced significant errors, including hallucinating contract values and inflating contracts worth $35,000 to appear worth $34 million. The AI flagged over 2,000 contracts as 'munchable' without understanding VA operations, legal requirements, or which services were essential for veteran care. At least 24 contracts identified by the system have been canceled, including contracts for maintaining gene sequencing devices for cancer treatment development, blood sample analysis for VA research, and tools to improve nursing care. The VA has canceled nearly 600 contracts overall as part of broader cost-cutting efforts. AI and procurement experts criticized the tool as fundamentally flawed and inappropriate for complex budgetary decisions affecting veterans' services.
Domain classification, causal taxonomy, severity scores, and national security assessments were LLM-classified and may contain errors.
AI systems that fail to perform reliably or effectively under varying conditions, exposing them to errors and failures that can have significant consequences, especially in critical applications or areas that require moral reasoning.
AI system
Due to a decision or action made by an AI system
Unintentional
Due to an unexpected outcome from pursuing a goal
Post-deployment
Occurring after the AI model has been trained and deployed