HISPI Project Cerebellum HISPI PROJECT CEREBELLUM

Purported DOGE Contract Review Tool Cited in Reports of AI-Driven Misjudgments in VA Budget Cuts

March 18, 2025

An alleged AI tool employing Language Model Models (LLMs) was deployed within the Veterans Affairs department to assess contracts for potential cuts based on minimal text analysis and simplified criteria. The system reportedly produced inaccurate values, incorrectly flagging vital healthcare services and research contracts for termination. At least two dozen such contracts were subsequently terminated.

Controversies like this underscore the importance of responsible AI governance and safe and secure AI practices. For those interested in shaping Project Cerebellum's efforts towards ensuring trustworthy AI, we invite you to JOIN US here.

This incident serves as an example of how HISPI Project Cerebellum TAIM can Map AI Incidents and work towards preventing harm and enforcing guardrails for AI.

Matched TAIM controls

Suggested mapping from embedding similarity (not a formal assessment). Browse all TAIM controls

Alleged deployer
department-of-government-efficiency, doge, sahil-lavingia
Alleged developer
department-of-government-efficiency, doge, sahil-lavingia
Alleged harmed parties
department-of-veterans-affairs-(va), veterans, veterans-receiving-care-through-the-va, va-clinical-and-research-staff, va-contractors

Source

Data from the AI Incident Database (AIID). Cite this incident: https://incidentdatabase.ai/cite/1103

Data source

Incident data is from the AI Incident Database (AIID).

When citing the database as a whole, please use:

McGregor, S. (2021) Preventing Repeated Real World AI Failures by Cataloging Incidents: The AI Incident Database. In Proceedings of the Thirty-Third Annual Conference on Innovative Applications of Artificial Intelligence (IAAI-21). Virtual Conference.

Pre-print on arXiv · Database snapshots & citation guide

We use weekly snapshots of the AIID for stable reference. For the official suggested citation of a specific incident, use the “Cite this incident” link on each incident page.