Bank of England’s Oversight of AI Regulations for Autonomous AI in Financial Services

The Bank of England is currently reassessing whether existing regulations are adequate to address the use of agentic AI in the financial sector, encompassing areas such as payments, trading, cybersecurity, and operations.

During the European Central Bank Forum on central banking in Portugal, Deputy Governor Sarah Breeden highlighted that current regulatory frameworks were not specifically designed to govern AI agents capable of autonomous action without direct human input. She emphasized the impracticality of relying solely on human supervision for every action taken by these systems.

Breeden pointed out that the existing frameworks were not constructed to anticipate autonomous agents operating in payments, trading, and operational capacities.

Agentic AI in Financial Workflows

Agentic AI refers to systems that have the ability to independently make decisions and execute tasks. In the financial realm, such systems are already in use for tasks like product recommendations, operational workflows, and trading activities.

Unlike traditional automated trading tools, agentic systems require less direct human oversight as they can pursue objectives and make decisions autonomously. Breeden noted that these systems could exhibit similar behaviors if trained on comparable data or designed with similar objectives in mind.

Recent advancements in AI models for identifying cyber vulnerabilities indicate a significant change in capabilities. Breeden highlighted that agentic AI systems can perform sequences of actions at scale and speed.

A report from the 2026 Cambridge Centre for Alternative Finance revealed that 81% of surveyed financial services firms are incorporating AI to some extent, with 52% actively adopting agentic AI. The report indicated that current usage primarily focuses on internal functions such as process automation, data visualization, software engineering, and knowledge management, with limited trading applications in lower-risk operational tasks.

Bank of England’s Concerns on Cyber Resilience Risks

Breeden emphasized cyber resilience as a critical concern for the Bank of England regarding the implementation of agentic AI. She noted a significant advancement in the cyber capabilities of the technology and highlighted the need for supervisors to assess risks across the entire financial system, not just individual firms.

While AI tools can enhance cyber defenses when employed by security teams, Breeden cautioned that there is a risk of these tools being exploited by malicious actors, potentially leading to attacks that could jeopardize financial stability.

Breeden also highlighted that open-source AI models may lag behind advanced closed models by a few months, posing challenges despite restrictions on releasing certain advanced models.

The IMF has warned that AI-enabled cyber risks should be treated as a financial stability issue due to the rapid scalability of attacks across sectors sharing digital infrastructure, potentially causing widespread disruption if multiple institutions are targeted simultaneously.

In response to these risks, Breeden suggested that authorities should assess the impact of simultaneous disruptions across multiple firms and conduct stress tests to prepare for such events. She also proposed stronger recovery requirements for core systems, including the option for one bank to assume another’s basic functions during an outage.

Other considerations include arrangements to ensure critical services continue in the event of compromised core systems, as well as the ability for key firms to quickly rebuild compromised systems.

Tobias Adrian from the IMF echoed concerns about AI’s impact on cyber resilience, particularly in shared digital infrastructure that could lead to correlated failures if targeted by malicious actors.

Regulatory Measures for Market Safeguards

Regulators are exploring guardrails, circuit breakers, and kill switches to limit or halt trading activities across markets in the event of severe disruptions caused by faulty AI models. Breeden warned that autonomous systems could exacerbate volatility if they respond similarly to market signals, especially if their objectives deviate from their intended purpose or public policy goals.

While the Bank of England previously believed existing rules were sufficient to manage AI-related risks, Breeden acknowledged that recent developments have exposed gaps in current frameworks.

Global Regulatory Review of AI Safeguards

The Financial Stability Board highlighted the unique challenge posed by AI agents in terms of human oversight and called for enhanced safeguards. In a June consultation, the FSB outlined 12 recommended practices for responsible AI adoption by financial institutions, covering governance, risk management, and addressing AI-related cyber and third-party risks.

The FSB emphasized that these practices are not intended to establish an international standard but suggested that firms define clear roles and responsibilities when utilizing AI, especially in critical functions.

Breeden emphasized the Bank of England’s focus on ensuring the resilience of financial firms as autonomous systems become more prevalent, encompassing controls at the firm level and safeguards across the market.

Explore more about AI and big data from industry experts at the AI & Big Data Expo event in Amsterdam, California, and London. Learn about this comprehensive event hosted as part of TechEx and co-located with other leading technology events here.

Powered by TechForge Media, AI News brings you the latest insights on enterprise technology events and webinars. Discover upcoming events and webinars in the tech industry here.