Agentic AI in Finance: Emerging Risks and the Regulatory Gap

Abstract

Financial institutions are deploying AI agents that plan and execute multistep workflows across research, portfolio rebalancing, trading, and client service. Fully autonomous use remains rare, creating a window in which governance can precede scale. This article gives portfolio managers, risk officers, boards, and supervisors a decision framework for that task. It classifies five emerging risks by evidence status and portfolio effect, including the possibility that common foundation models create behavior not visible in conventional exposures before stress. It compares US, Canadian, and EU regulatory approaches and shows how divergence creates compliance, competitive, and cross-border risks. It then proposes four controls: autonomy-proportional governance, security by design, agent stress testing, and fiduciary controls. A stylized asset manager case and an eight-question decision gate show how institutions can inventory agents, detect weak human review, constrain execution, test model updates and common-provider scenarios, and document client obligations before granting greater autonomy.