UK financial firms rely heavily on third-party AI, but many say they lack deep knowledge of how those systems work. A new Resultsense analysis finds that the duties around supplier AI are fragmented, while the means of checking it remain patchy.
Watch Desk analysis
What happened
The analysis cites a Bank of England and FCA survey in which third-party systems made up 53% of high-materiality AI reported by UK financial firms in 2026, up from 35% in 2024. Only 24% of respondents said they relied on deep knowledge of supplier systems, compared with 77% for systems built in-house.
The gap continues after procurement. Around half of respondents reported ongoing monitoring, leakage testing or periodic review. Resultsense says the relevant UK expectations vary: data-protection rules provide a firmer basis for due diligence and checks where personal data is involved, while other requirements are voluntary, proposed or focused on cyber security rather than AI decision-making.
Why it matters
Buying an AI system does not make its risks someone else’s problem. But contracts and vendor assurances may not give a buyer enough access to understand what a system does or to check how it changes after deployment. Resultsense says limited transparency into third-party systems was a major constraint for 34% of survey respondents.
The figures describe 196 firms in a regulated sector, not UK businesses generally. Even within that group, the picture is not a simple absence of oversight: the analysis says vendor due diligence is common, while monitoring after deployment is less consistent.
Our read
The awkward question is not whether firms have signed a supplier questionnaire. It is whether they can show what happens when a model changes, a problem appears or a supplier will not open the bonnet. A contract can allocate responsibility; it cannot, by itself, make a system transparent.
What to watch
- Whether the next Bank and FCA survey shows firms gaining deeper knowledge of supplier systems.
- Whether revised ICO guidance gives buyers practical examples for AI supplied by third parties.
- Whether the Cyber Security and Resilience Bill leads to AI providers being designated as critical suppliers.
Discussion spark: Should firms be required to prove they can monitor supplier AI after deployment, even if that means demanding access providers may resist?
Sources and evidence
- Can UK firms prove they govern the AI they buy? – Resultsense (2 November 2026, 09:37 UTC)
- PlanAssist AI planning tool moves to 15 English councils – Resultsense (2 November 2026, 09:37 UTC)
- Source update (2 November 2026, 09:37 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.