Inspector General Flags Missing Protections in SBA's AI Fraud Detection System
The Small Business Administration deployed artificial intelligence software from Palantir to identify fraud in pandemic-era loans without properly classifying it as a high-impact use case requiring federal risk management standards. The agency failed to list the fraud detection pilot on its public AI inventory, though it had publicly acknowledged using the tool to investigate PPP and Economic Injury Disaster Loan fraud. The SBA's Office of Inspector General determined these oversight gaps left the program operating without necessary safeguards.
The Small Business Administration began using Palantir's artificial intelligence software to investigate fraud across pandemic-era lending programs, including the Paycheck Protection Program and Economic Injury Disaster Loans. In February, the agency publicly announced it had suspended over 111,000 California borrowers suspected of fraud using this partnership. However, the SBA did not formally designate this fraud-detection pilot as a "high-impact use case" under federal standards—a classification that would have triggered mandatory safeguards.
Federal guidance issued in April 2025 established that AI systems affecting access to critical services like loans should be treated as high-impact and subjected to risk assessments, appeals processes, and human oversight. The Inspector General's investigation found the SBA lacked documentation justifying its decision to exclude the Palantir tool from this classification, and the agency had not established the AI governance board required by oversight rules.
This oversight gap could affect thousands of small business borrowers whose loan status was determined partly by an AI system operating without federally mandated safeguards. Affected borrowers may lack clear appeal mechanisms or transparency about how the tool flagged their applications as fraudulent. The case illustrates potential tensions between rapid deployment of fraud-detection technology and accountability standards, raising questions about whether similar gaps exist across other federal agencies using AI for high-stakes decisions affecting citizen eligibility for benefits and services.