SBA Used Unvetted AI to Target Loan Fraud

The Small Business Administration failed to meet federal risk standards while using AI to flag pandemic loan fraud.

Updated on Sept. 29, 2026 in Artificial Intelligence

Bold flat-color editorial illustration showing a heavy concrete block suspended by steel chains, representing institutional lack of transparency. — The Small Business Administration bypassed federal oversight and transparency requirements while deploying AI software to identify fraud in pandemic-era small business loans. AI Illustration. Upload story photo >

Live Poll

Should federal agencies face stricter oversight when using AI to manage public benefits and programs?

The Small Business Administration utilized Palantir software to identify suspected fraud in COVID-19 pandemic assistance loans without adhering to mandatory federal risk management standards. The agency omitted the program from its public AI use case inventory and failed to establish an AI governance board.

Why it matters

The Office of Management and Budget mandates specific risk mitigation for AI tools affecting federal benefits to ensure transparency and fairness. By failing to categorize its pilot as a high-impact use case, the SBA bypassed oversight requirements intended to protect applicants.

The SBA utilized Palantir AI software to surface leads for criminal enforcement, but the agency has not yet convened an AI governance board or completed mandated impact assessments. The OIG issued 5 recommendations to the agency to rectify these lapses.

The players

Small Business Administration

This is a United States federal agency that provides support to entrepreneurs and small businesses.

Office of Inspector General

This office conducts independent audits and investigations to promote economy and efficiency within federal agencies.

Palantir

This technology company specializes in big data analytics and software for the public and private sectors.

Office of Management and Budget

This cabinet-level office oversees the implementation of the President's vision across the executive branch.

The details

The AI model was deployed to detect fraud in pandemic relief funds, resulting in the suspension of 111,620 California borrowers. The agency failed to provide consistent remedies for those affected by the automated flagging process while it simultaneously expanded its use of the technology.

April 2025: The OMB released a memorandum requiring specific AI risk mitigation steps for federal agencies.
January 2026: The SBA signed a contract with Palantir for a fraud detection pilot and bootcamp.
February 2026: The SBA suspended 111,620 California borrowers for suspected loan fraud.
April 2026: The OIG completed its formal review of the SBA fraud prevention program.
July 2026: The SBA formalized the use of Palantir software for ongoing fraud prevention.

The Tech Race

The agency's failure to label its tool as a high-impact use case contrasts with the strict oversight required by the April 2025 Office of Management and Budget AI risk mitigation memorandum. This incident underscores the tension between rapid AI deployment for fraud detection and the necessity of federal accountability standards.

Borrowers who were suspended face significant financial disruption due to the lack of consistent remedies or assessment processes in the SBA's automated system. Individuals affected by these AI-driven decisions currently have limited clarity on how to appeal their status.

The takeaway

Automated fraud detection systems require rigorous human oversight and transparent governance to prevent systemic errors that impact thousands of citizens. Agencies must ensure that efficiency-focused technologies remain strictly compliant with federal risk management mandates.

What happens next

The House of Representatives has passed legislation that requires the SBA to provide a formal report to Congress regarding its specific AI use cases and governance structures.