Safeguarding Federal Payment Infrastructure: Trustworthy AI for Improper-Payment Prevention, Synthetic Identity Detection, and Public-Benefit Disbursement Integrity in the United States

Abstract

Federal payment integrity has become a high-stakes governance problem in the United States because public-benefit disbursement systems must now operate at digital scale while defending against increasingly adaptive fraud, identity manipulation, and data-quality failures. Recent federal estimates show that improper payments remain material across major programmes, while agencies and oversight bodies have increasingly explored data matching, risk analytics, and automated screening to strengthen prevention and recovery. Yet the policy challenge is not simply whether artificial intelligence can detect more anomalies. In public disbursement settings, analytical performance must be reconciled with due process, transparency, data governance, and the need to avoid delaying legitimate payments to eligible recipients. This paper develops a literature-based conceptual framework for trustworthy AI in federal payment infrastructure, with particular attention to improper-payment prevention, synthetic identity detection, and public-benefit disbursement integrity. Using an integrative review of peer-reviewed scholarship and authoritative U.S. government sources published, the paper synthesises four literatures that are often treated separately: payment integrity, digital identity assurance, AI-enabled fraud analytics, and public-sector AI governance. The paper proposes a multilayer framework that combines identity-proofing controls, multimodal risk scoring, graph-based network analysis, human-centred adjudication, and continuous governance monitoring. Its central argument is that effective federal payment integrity requires not only predictive accuracy but institutional trustworthiness: auditable models, calibrated thresholds, fairness safeguards, contestability, and shared cross-programme infrastructure. The paper contributes a structured analytical model, implementation logic, policy implications, and an evaluation template for future empirical testing.