Bank of England: Agentic AI Redefines Finance

The Bank of England Faces Agentic AI in Finance

The Bank of England is currently assessing whether existing regulations are sufficient to govern the use of agentic artificial intelligence in the financial sector. This technology, which includes systems capable of operating without direct human intervention, is already present in areas such as payments, trading, cybersecurity, and financial operations. Sarah Breeden, the Bank's Deputy Governor, expressed her concerns at the European Central Bank Forum in Portugal. She emphasized that current regulatory frameworks were not designed to manage autonomous agents capable of making independent decisions in these key sectors.

Breeden insisted that human oversight cannot be the solution for every action taken by these systems. Regulatory frameworks must evolve to account for the capabilities of autonomous agents in financial operations. She highlighted that relying on constant human supervision is impractical, given the independent nature of these systems.

The Rise of Agentic AI in Financial Processes

Agentic AI is distinguished by its ability to make decisions and execute tasks autonomously. In the financial sector, these systems are already integrated into processes such as product recommendations and trading operations. Unlike traditional automated trading tools, these systems can pursue objectives independently, requiring less human supervision.

Sarah Breeden pointed out recent advancements in AI models, particularly in identifying cybersecurity vulnerabilities. These systems can chain actions at scale and speed, representing a significant shift in their capabilities. She explained that these systems, if trained on similar data or designed around similar objectives, can act in similar ways.

A report from the Cambridge Centre for Alternative Finance in 2026 reveals that 81% of financial services firms are using AI to varying degrees, with 52% of them already adopting agentic AI. However, the majority of current applications focus on internal functions such as process automation and knowledge management, with trading use still limited to low-risk tasks.

Cyber Resilience Risks According to the BoE

Cyber resilience is a major concern for the Bank of England regarding agentic AI. Breeden described a "paradigm shift" in cyber capabilities, necessitating a risk approach at the level of the financial system rather than individual firms.

AI tools can bolster cyber defenses, but they also pose risks if they fall into the wrong hands. Breeden highlighted that open-source models could quickly catch up to the most advanced closed models, which worries authorities despite restrictions on the publication of certain models.

The IMF has also warned about the cyber risks associated with AI, emphasizing that attacks can spread rapidly and affect multiple institutions simultaneously. Breeden suggested that authorities should prepare for massive disruptions rather than isolated interruptions, considering stress tests to assess the potential impact of such situations.

The Bank of England is contemplating stricter recovery requirements for essential systems, such as allowing one bank to take over the functions of another in the event of a failure. Other options include establishing backup systems to ensure the continuity of critical services.

Market Protection Measures Under Discussion

Regulators are also examining protective mechanisms to limit disruptions caused by faulty AI models. Breeden mentioned the importance of safeguards, circuit breakers, and kill switches to control trading in the markets.

She warned that autonomous systems could amplify volatility if they react similarly to the same market signals, especially if their objectives diverge from initial goals or public policies.

The Bank of England acknowledged that current rules seemed adequate to manage AI-related risks, but recent developments have revealed gaps in these frameworks.

A Global Perspective on AI Regulation

The Financial Stability Board recently highlighted the challenges posed by AI agents for human oversight, calling for stricter protective measures. In June, the FSB proposed 12 practices for the responsible adoption of AI in the financial sector, covering governance, risk management, and cybersecurity issues.

These practices are not intended to establish a binding international standard but to encourage firms to define clear roles when using AI, particularly in critical functions.

Sarah Breeden concluded by stating that the Bank of England's goal is to ensure the resilience of financial firms as autonomous systems become more widespread. A review of controls at the firm level and market-wide protective measures is underway to ensure that these systems do not compromise financial stability.