Financial regulators are grappling with an uncomfortable reality: the rulebooks they rely on were written for a different era. As autonomous artificial intelligence systems gain prominence across banking and investment operations, the Bank of England's deputy governor for financial stability Sarah Breeden has become the latest senior policymaker to flag critical gaps in how we monitor and control these rapidly evolving technologies. Speaking at the European Central Bank Forum on central banking in Portugal on Tuesday, Breeden articulated a concern that is gaining traction among global supervisors: existing governance structures may be fundamentally unsuited to the challenge posed by AI agents that can make decisions and execute transactions with minimal human intervention.
The challenge Breeden highlighted touches on a core tension in modern finance. Regulators have long insisted on the "human in the loop" principle—the idea that critical decisions affecting the financial system should involve people who can be held accountable. Yet as artificial intelligence becomes more capable and more embedded in trading systems, risk management algorithms, and transaction processing, the practical reality is that humans cannot feasibly review every action. The sheer volume and velocity of decisions that autonomous systems can make overwhelms traditional oversight approaches. Financial institutions increasingly rely on AI to spot patterns, execute trades, and manage complex portfolios. The regulatory frameworks that evolved to govern human traders and automated systems with humans supervising every decision simply do not contemplate truly autonomous agents making high-stakes financial decisions with only sporadic human review.
Breeden's intervention carries particular weight because financial stability lies at the heart of central banking. Her warnings about governance gaps suggest that senior officials at major institutions like the Bank of England see autonomous AI not merely as a technology management issue but as a systemic risk question. When an autonomous system makes decisions affecting financial markets without continuous human oversight, the potential for rapid, widespread losses multiplies. Unlike a single trader who can be stopped, an AI agent might execute thousands of transactions in seconds before anyone realizes something has gone wrong.
The broader regulatory landscape reinforces the urgency of Breeden's call for reform. The Financial Stability Board, which coordinates policy among the world's major financial regulators, issued specific warnings in June emphasizing that AI agents pose a "distinct challenge to human oversight." This language is telling—it acknowledges that autonomous AI is not simply a more sophisticated version of existing automation, but represents a qualitatively different phenomenon requiring fresh thinking about how to maintain control and accountability.
The cybersecurity dimension adds another layer of complexity. Analysts have flagged that autonomous AI agents deployed across banking systems could create new vulnerabilities that malicious actors might exploit. A compromised AI system operating autonomously could potentially cause damage before detection and intervention become possible. The interconnected nature of modern finance means that problems at one institution could rapidly cascade, particularly if multiple firms rely on similar AI systems or if those systems interact in ways that amplify instability.
Southeast Asia and Malaysia, though geographically distant from the Bank of England, cannot treat this as a distant regulatory concern. Malaysian banks increasingly adopt AI and automation in their operations, and many are integrated into global financial networks. When major economies like the United Kingdom and the eurozone begin tightening rules around autonomous AI, those changes ripple outward. Malaysian financial institutions that do business internationally will need to comply with stricter standards or modify systems serving international operations. Moreover, as regulators worldwide converge on new standards for AI governance, there will be pressure on domestic supervisors to adopt similar frameworks.
The call for "more sophisticated governance and accountability frameworks" implies significant structural changes ahead. This could mean new requirements for testing autonomous systems before deployment, stronger audit trails for AI decisions, enhanced monitoring systems to detect anomalies in real time, and clearer rules about who bears responsibility when an autonomous system causes losses. Financial institutions may need to invest substantially in compliance infrastructure, which could raise costs and reshape competitive dynamics.
Breeden's framing also subtly challenges the assumptions behind rapid AI adoption in finance. While technology vendors and some financial firms have promoted autonomous AI as a path to efficiency and better risk management, senior regulators are signalling caution. They are asking hard questions about whether the benefits justify the systemic risks, and whether existing governance approaches can adequately manage these tools. This regulatory skepticism may slow deployment in some areas while accelerating it in others where controls can be more clearly established.
The tension between innovation and stability that Breeden's remarks highlight will likely define financial regulation throughout the remainder of this decade. Financial institutions want to harness AI's productivity benefits, but regulators must ensure that the pursuit of efficiency does not create hidden risks that could destabilize markets or harm consumers. Breeden's intervention suggests that the Bank of England and its peers globally are moving from passive observation toward active policy development, recognizing that leaving autonomous AI systems to operate under outdated rules is not a sustainable position.
