AI in action · Regulation
86% of UK Financial Firms Now Use AI, BoE and FCA Find
The fourth Bank of England and FCA survey finds 86% of UK financial firms use AI, up from 75% in 2024. Only 2% report AI agents with broad autonomy.
Published on Monday, October 26, the fourth survey of artificial intelligence in UK financial services by the Bank of England and the Financial Conduct Authority puts AI use at 86% of the 196 responding firms, up from 75% in 2024. The regulators caution that the two samples differ, so the comparison is approximate. The mix has shifted too: generative AI now accounts for 56% of the AI systems firms reported, where foundation models made up 17% two years ago.
Agents are being deployed with the brakes on. Most firms with AI agents in production use them only to support human decisions, and just 2% report giving them broad autonomy. Agents that write to customers automatically are in use at 16% of firms, rising to 34% when planned projects are counted, but no more than 9% expect agents to approve customer outcomes, execute trades, initiate payments or change limits.
The share of systems rated highly material fell to about 7% from 16% in 2024, although their absolute number rose, and the top ten respondents account for 43% of all reported systems. Governance looks solid on paper: 94% of firms have a firmwide AI policy and more than 90% use human approval, pre-deployment review and vendor due diligence. Ongoing monitoring, leakage testing and periodic review, the controls that catch problems after launch, are used by only around half.
The concentration figures matter most for financial stability. Third-party systems now make up 39% of all reported AI and 53% of the highly material ones, up from 35% in 2024. The top three cloud and compute providers account for 79% of firms' relationships and the top three model providers for 54%. For the systems they build themselves, 77% of firms rely on deep in-house knowledge to understand them; for bought-in systems only 24% can, and 81% fall back on vendor assurances and contracts.
Firms expect the risks to grow. Data protection and security, supply-chain dependence and generative-AI faults such as hallucinations and prompt injection top the list, and 88% expect frontier AI's cyber capabilities to have a moderate or high impact on their risk profile within two years. Taken with the Bank's reported interest in an AI scenario for its next system-wide stress test, the survey sketches where supervisors will look next: not at whether firms use AI, but at how few suppliers the whole sector now depends on.
Sources: Bank of England and FCA — Artificial intelligence in UK financial services – 2026 (Oct 26, 2026); Regulation Tomorrow — BoE and FCA publish joint report on AI in UK financial services (Oct 2026)