Generative Artificial Intelligence Governance in Financial Services: Managing Hallucination, Bias, and Compliance Risks

Abstract

Generative artificial intelligence has transitioned from experimental technology to embedded operational infrastructure across financial services, deployed for marketing, customer communications, anti‑money laundering monitoring, and compliance functions. Yet the same capabilities that drive efficiency introduce novel risks that existing risk management frameworks are ill‑equipped to address. This article examines the governance challenges posed by generative AI in regulated financial institutions, focusing on three risk categories identified as priorities by financial regulators: hallucinations and inaccurate outputs, algorithmic bias and concept drift, and the emerging autonomy of AI agents. The analysis draws on the Financial Industry Regulatory Authority's 2026 Regulatory Oversight Report, the European Banking Authority's implementation of the EU AI Act, the Monetary Authority of Singapore's proposed AI risk management guidelines, and the FINOS AI Governance Framework. Findings indicate that existing securities laws and supervisory rules apply with equal force to generative AI‑powered operations, yet compliance gaps persist due to the novelty of the technology and the absence of standardised testing protocols. Persistent challenges include the difficulty of verifying output accuracy, the opacity of model decision‑making, and the accountability vacuum created by autonomous AI agents. The analysis concludes by outlining a multi‑layered governance framework encompassing cross‑functional oversight, usage policies, testing and monitoring protocols, and recordkeeping practices tailored to generative AI.