UK regulators dominated this week’s finance AI agenda. The Bank of England and FCA published their 2026 survey showing near-universal AI adoption, and the FCA chief warned that AI is amplifying financial crime and market risks. The common thread is a shift from experimentation to autonomous, agentic deployment, which is pushing regulators toward incident reporting, stress-testing and accountability expectations.
Top story: The Bank of England and FCA’s joint survey shows 86% of UK financial firms now use AI, with 80% deploying AI agents but only 2% granting broad autonomy.
BoE and FCA survey: 86% of UK finance firms use AI
Regulation Tomorrow (Hogan Lovells) · Regulation
The Bank of England and FCA published their fourth joint survey on 26 October. It shows AI adoption at 86% of firms, up from 75% in 2024. Commentary on the findings notes that 80 percent of respondents have AI agents deployed, yet only 2 percent report any agent with broad autonomy, so governance of those limits is the next supervisory question.
FCA chief: AI puts market abuse and crime ‘on steroids’
Tech Times · Risk
Speaking at Mansion House on 22 October, FCA chief executive Nikhil Rathi said AI is intensifying four threats: market volatility, concentration risk from dominant cloud and AI providers, novel market abuse, and cross-border financial crime. The FCA will also use AI itself, positioning as an “agentic regulator”. Firms should expect tougher supervisory scrutiny of both AI-enabled threats and their reliance on a few hyperscalers.
FCA tells banks to report AI failures and near-misses
Bitdefender Hot for Security · Regulation
The FCA wants firms to report significant AI failures and near-misses, not just incidents that cause actual damage. It also urges firms to stress-test AI systems, maintain human oversight and prepare fallback procedures. The aviation-style safety reporting model could become a template for compliance teams building AI incident processes.
https://www.bitdefender.com/en-us/blog/hotforsecurity/uk-watchdog-banks-ai-failures
Bank of England weighs AI-driven market shock for stress test
Resultsense (via Bloomberg/Claims Journal) · Finance
The Bank of England is discussing AI-caused turmoil as a possible theme for its next system-wide exploratory scenario, though nothing has been decided. The concern is that firms using similar models from a handful of vendors could behave in correlated ways in fast markets. Trading desks, asset managers and insurers should consider how concentrated their model dependencies are.
https://www.resultsense.com/news/2026-10-21-bank-of-england-ai-stress-test-option/
EU AI Act high-risk deadline for banks pushed to December 2027
CMS Law · Strategy
Following the Digital Omnibus political agreement, high-risk rules will begin to apply to standalone AI systems from 2 December 2027. This matters for credit scoring and similar banking and insurance use cases. The agreement also sets exceptions where national authorities, not the AI Office, will oversee AI systems built on general-purpose models in financial institutions when the same provider develops both model and system. The extra time lets EU-facing firms build governance, but the underlying obligations remain.