The Bank of England convened three roundtable discussions in late 2025 with representatives from regulated firms to explore the responsible adoption of artificial intelligence and machine learning. Published on 16 February 2026, the summary captures industry views on constraints and the Bank’s potential role in supporting innovation while maintaining stability.
Participants expressed broad support for the Prudential Regulation Authority’s existing framework. They praised its principles-based, outcomes-focused approach, which provides clear guardrails without stifling progress. Supervisory Statement 1/23 on model risk management received particular mention as pragmatic and enabling for responsible AI use.
Most attendees saw no immediate need for detailed AI-specific rules or a dedicated Bank or PRA AI sandbox. They viewed the Financial Conduct Authority’s Supercharged Sandbox and AI Live Testing initiatives as sufficient for experimentation.
The conversations align with the Bank’s dual mandate: safeguarding financial stability and promoting growth. By engaging directly, the Bank aims to identify barriers that slow adoption and consider ways to ease them responsibly.
Firms raised practical challenges in scaling AI, from data quality and governance to integration with legacy systems. Yet the tone remained constructive. Regulated entities want clarity and support that encourages safe experimentation rather than prescriptive mandates.
This engagement matters for the wider sector. As banks and insurers embed AI in credit decisions, fraud detection, and customer service, regulators must balance innovation with risk control. The roundtables signal proactive dialogue that could shape future guidance.
What happens if constraints persist without targeted support? Adoption may lag in the UK compared with other markets. The Bank’s approach seeks to avoid that outcome by listening closely to practitioners and refining its stance accordingly. The published summary offers a transparent view into those early discussions.
Author:Oje. Ese