BoE Finds AI Surges in UK Finance Amid Sovereign Tech Fears
AI deployment across British finance is accelerating swiftly, revealed by an investigation from the Bank of England (BoE).
The joint research project with the Financial Conduct Authority, entitled ‘Artificial intelligence in UK financial services - 2026’, established that adoption climbed to 86% across participants, contrasting with 75% in 2024.
Every domestic and multinational banking institution surveyed recorded operational implementation, alongside 93% of insurance businesses.
Sector participation in machine learning initiatives is near-universal: 94% maintain working models live in production, climbing to 97% once trial programmes or proofs of concept are tallied, and hitting 99% when incorporating twelve-month pipeline schedules.
Deployment continues to be uneven across departments, however. Though utilised in various corporate units, 31% of live deployments sit within operations and IT.
James Maxfield, Chief Client Officer at London-headquartered tech firm Duco, highlights: “The report shows operations is one of the areas where firms most expect AI to grow, with seven in ten using it or planning to within the next year.”
- AI adoption across UK financial services rose to 86% in 2026, up from 75% in 2024, according to the joint Bank of England/FCA report
- 94% of surveyed firms have machine learning models live in production, rising to 97% when including pilots/POCs, and 99% when factoring in 12-month pipeline plans
- Only 2% of organizations permit fully unconstrained AI autonomy, while 46% allow constrained execution and 22% allow conditional independence — most of the remainder use AI only for guidance/decision support (around 8 in 10 firms)
- Cybersecurity risk concerns are rising sharply: 57% of firms rate their current cyber risk as severe/elevated, a figure expected to climb to 80% within three years
- 79% of all logged hosting/compute dependencies are tied to the three major hyperscalers (Google, Microsoft, Amazon), with 98% of businesses using them — yet only 35% of high-risk infrastructure ties have tested migration exit plans
Proliferation of automated decision tools
Organisations have embraced autonomous programmes in volume, but the central bank notes that overall delegation stays restricted.
Eight in 10 participants employ agent architectures for guidance or decision support.
By comparison, 46% authorise constrained execution, 22% provide conditional independence and a mere 2% permit unconstrained autonomy.
Systems granted wider operational latitude are grouped mainly in Risk and Compliance, Operations and IT, alongside Sales and Marketing.
Concurrently, overall post-launch supervision shows deficits. James adds that firms are thorough about approving AI before it goes live, but only around half keep monitoring it afterwards.
“The ongoing assurance that it keeps doing the right job will come from the people who know the work best,” he says.
Balancing efficiency gains and operational hazards
Primary drivers cited by organisations include productivity boosts, task automation and departmental efficiency, trailed by operational resilience alongside cybersecurity, risk oversight and control rigour.
Conversely, cyber risk inflation forms the leading operational worry: 57% categorised their vulnerability as severe or elevated today, which swells to 80% across a three-year timeframe.
Third-party vendor exposure followed closely, where acute worry climbs from 40% currently to 63% across one to three years.
Heavy hosting consolidation mirrors this apprehension.
European and British bodies increasingly voice caution surrounding over-reliance upon American technology behemoths, prompting strong domestic momentum towards self-governing computational infrastructure.
Without [more sovereign compute and cloud], we risk losing our status as a leading global AI ecosystem, as well as a leader of wider industries.”
Cloud exposure and American computing monopolies
Sovereign capability remains heavily compromised by current partnerships.
Data sourced via Swiss company Proton reveals that over 74% of European listed corporations depend on US tech services.
Within the central bank's findings, 79% of all logged hosting or compute dependencies link directly to the three prominent hyper-scalers (evidently US tech groups Google, Microsoft and Amazon), which 98% of businesses engage.
Out of 169 registered infrastructure ties, 37% present both high criticality alongside severe migration complexity.
Merely 35% among that vulnerable segment have formally tested migration exit pathways.
Westminster expands domestic tech backing
To bridge the divide against US and Chinese dominance, Britain's leadership is directing increased state capital towards homegrown computing.
UK Prime Minister Andy Burnham recently allocated £60m (US$80m) to the Sovereign AI Fund for equity investments in high-potential domestic technology ventures, expanding upon the £500m (US$671m) committed under Keir Starmer’s premiership.
Simultaneously, the administration pledged £60m (US$80m) towards an R&D Procurement Scheme alongside £18m (US$24m) to fund a UK Advanced Chip Design Academy, aiming to offset foreign reliance on international semiconductor fabrication.
Infrastructure autonomy across British banking
Defining digital sovereignty varies, as market participants frequently prioritise regulatory compliance and domestic hosting borders ahead of total isolation.
For instance, Starling Bank stands as the inaugural domestic lender to build an AI system using UK sovereign cloud services, aspiring to inspire similar ecosystem collaboration.
Teaming up with NVIDIA, the lender's TFM model trains on Civo sovereign architecture to power its proprietary tooling, including a pioneering domestic agentic assistant.
Harriet Rees, Group CIO of Starling Bank and a UK Government AI Champion for Financial Services, previously told AI Magazine: “Without [more sovereign compute and cloud], we risk losing our status as a leading global AI ecosystem, as well as a leader of wider industries.”
Bank of England's key partners:
Accenture: Accenture collaborates with the Bank of England to modernise its core Real-Time Gross Settlement infrastructure, enhancing national wholesale payment resilience.
CEO: Julie Sweet | Headquarters: Dublin, Ireland
De La Rue: De La Rue partners with the Bank of England to design, print, and secure the United Kingdom’s physical polymer banknotes.
CEO: Clive Vacher | Headquarters: Basingstoke, Hampshire, United Kingdom
Swift: Swift collaborates with the Bank of England to provide secure global financial messaging rails and implement ISO 20022 messaging standards.
CEO: Javier Pérez-Tasso | Headquarters: La Hulpe, Walloon Brabant, Belgium
Company portals
Executives
Harriet Rees
UK Government AI Champion - Financial Services
James Maxfield
Chief Product Officer
