Bank of England and FCA survey: 80% of UK finance firms run AI agents, only 2% with broad autonomy
A joint Bank of England and Financial Conduct Authority (FCA) report published on 26 October 2026 surveyed 196 UK financial services firms, 105 of which completed the full survey including AI agent questions. 80% of full-survey respondents have AI agents deployed, and only 2% run broadly autonomous agents with limited real-time human involvement. Overall AI use stands at 86% of firms, up from 75% in 2024 and 58% in 2022. The regulators flag cyber risk amplification as the most significant systemic concern. They also point to growing dependence on third-party AI and cloud providers. The report is a sector adoption and risk survey, not an incident or threat report.
Detalhes técnicos
- Agent autonomy: 80% use agents for decision support, 46% for executing actions within predefined rules, 22% for conditional autonomy with human approval, and 2% for broad autonomy.
- Sensitive uses such as executing trades, initiating payments or approving customer outcomes are expected by no more than 9% of firms, even counting planned deployments.
- Third-party implementations make up 39% of AI systems (33% in 2024) and 53% of systems classed as highly material (35% two years earlier).
- The three largest cloud and compute providers account for 79% of reported relationships.
- 40% of respondents cite information security and operational resilience as a major constraint on third-party AI adoption, and 34% cite limited transparency into external AI systems.
- 57% rate the current risk of AI-related cyber threats being exploited at scale as high or severe, rising to 80% over the next one to three years.
- 94% have firmwide AI policies, but ongoing monitoring, leakage testing and periodic reviews are each reported by only about half.
Remediação
The article states no formal remediation. The findings point to gaps in AI monitoring, leakage testing, periodic system review, third-party transparency and concentration risk.
Linha do tempo
-
Bank of England and FCA survey: 80% of UK finance firms run AI agents, only 2% with broad autonomy
ler a análise desta fonte ocultar
A joint Bank of England and Financial Conduct Authority (FCA) report published on 26 October 2026 surveyed 196 UK financial services firms, 105 of which completed the full survey including AI agent questions. 80% of full-survey respondents have AI agents deployed, and only 2% run broadly autonomous agents with limited real-time human involvement. Overall AI use stands at 86% of firms, up from 75% in 2024 and 58% in 2022. The regulators flag cyber risk amplification as the most significant systemic concern. They also point to growing dependence on third-party AI and cloud providers. The report is a sector adoption and risk survey, not an incident or threat report.
Technical Details
- Agent autonomy: 80% use agents for decision support, 46% for executing actions within predefined rules, 22% for conditional autonomy with human approval, and 2% for broad autonomy.
- Sensitive uses such as executing trades, initiating payments or approving customer outcomes are expected by no more than 9% of firms, even counting planned deployments.
- Third-party implementations make up 39% of AI systems (33% in 2024) and 53% of systems classed as highly material (35% two years earlier).
- The three largest cloud and compute providers account for 79% of reported relationships.
- 40% of respondents cite information security and operational resilience as a major constraint on third-party AI adoption, and 34% cite limited transparency into external AI systems.
- 57% rate the current risk of AI-related cyber threats being exploited at scale as high or severe, rising to 80% over the next one to three years.
- 94% have firmwide AI policies, but ongoing monitoring, leakage testing and periodic reviews are each reported by only about half.
Remediation
The article states no formal remediation. The findings point to gaps in AI monitoring, leakage testing, periodic system review, third-party transparency and concentration risk.