Artificial intelligence is becoming more deeply integrated into financial services, from fraud detection and customer support to cybersecurity and internal operations. But as banks begin experimenting with more autonomous AI systems, regulators are asking a difficult question: how can financial institutions benefit from AI without losing control of critical decisions?
A Reuters report published on October 13 said banking watchdogs in the United States and Europe were increasing scrutiny of AI use amid concerns about cyber threats and financial stability.
The attention reflects a shift from treating AI as an optional productivity tool to recognising it as technology that may affect essential financial systems.
Why Banks Are Interested in AI Agents
Traditional AI applications often perform a specific task, such as classifying a transaction or summarising a document. More advanced AI agents can be designed to plan and carry out sequences of actions, potentially interacting with other software and systems along the way.
Banks may use these tools to support cybersecurity investigations, analyse suspicious transactions, assist employees and automate routine processes. In theory, faster analysis could help institutions identify problems and respond to threats more quickly.
However, greater autonomy also creates additional risks. An AI agent with excessive permissions could make an incorrect change, expose sensitive information or take an action that was not intended by its operators.
The concern is especially serious in financial services, where operational mistakes can affect customers, payments and market confidence.
What Regulators Are Watching
According to Reuters, US agencies have been asking banks about their deployment of agentic AI, including how these systems are controlled and audited. European authorities, including the Bank of England and European Central Bank, have also emphasised the importance of cyber resilience and recovery planning.
Regulators are concerned about whether banks can identify what an AI system has done, explain why it acted and intervene when its behaviour becomes unsafe.
These questions become more complicated when systems depend on external AI providers or interact with multiple internal platforms. Institutions must understand which information the system can access, what actions it is authorised to perform and how its decisions are reviewed.
The Importance of Human Oversight
Allowing AI to assist with a task is different from giving it unrestricted authority to execute that task.
Banks can reduce risk by limiting agent permissions, requiring human approval for high-impact actions and keeping detailed records of automated activity. They should also test AI systems under realistic conditions before deploying them in sensitive environments.
Independent backups, tested recovery procedures and clear incident-response plans remain essential. AI tools may help identify threats, but they should not become the only line of defence.
Financial institutions also need to evaluate how AI systems behave when they encounter incomplete information, malicious instructions or unexpected technical failures.
Innovation Without Losing Control
Stronger oversight does not necessarily mean banks must stop adopting AI. Instead, it can help them identify which applications are suitable for automation and which require tighter controls.
AI could improve efficiency, help employees process information and support faster responses to emerging threats. But financial institutions need evidence that these benefits outweigh the risks in each deployment.
The challenge for regulators is to encourage useful innovation without allowing poorly controlled systems to threaten customer data or financial stability.
As AI agents become more capable, trust will depend not only on what they can do but also on whether institutions can monitor, limit and reverse their actions when necessary.
