AI use and ownership in the financial sector, as supervisors see it

Every year the ECB collects data on the use of new technology at the banks it supervises. Pedro Machado, a member of the ECB’s Supervisory Board, reported in February 2026 that more than 85 per cent of large banks use AI in some form, and that the number rises every year. In Amsterdam in September 2026, the Board’s Chair, Claudia Buch, put the figure at more than 90 per cent, and 85 per cent for generative AI. Alongside the familiar uses in credit risk, fraud detection and transaction monitoring, AI is now also finding its way into IT operations, the analysis of legal documents and customer contact.

Governance is less advanced. In workshops with 13 banks using AI for credit scoring and fraud detection, described in the Supervision Newsletter of November 2025, about half had their own AI policy or an AI committee. According to the authors, the second and third lines of defence still need to monitor usage more closely, and they describe appointing a Chief AI Officer as an emerging practice. The ECB itself warns that the sample is small. In February Machado saw the same thing more broadly: responsibility is spread across IT, data teams, the business and the control functions, with no clear framework for who is accountable. In September Buch added that digital innovation needs the attention of the board and cannot be left to technical experts alone, and that the bank remains responsible for decisions reached with AI.

EIOPA, the European insurance supervisor, finds a similar picture among insurers. It asked 347 insurers in 25 countries about generative AI, as at the second quarter of 2025; the report appeared in February 2026. Sixty-five per cent already use it and another 23 per cent plan to within three years, although most applications are still at the pilot stage and almost two thirds concern internal processes. Forty-nine per cent have their own AI policy, twice as many as in the 2023 survey. A shortage of skilled staff is among the biggest obstacles, and for now at most insurers a person checks what the system produces.

In the Netherlands, the AFM (the Dutch Authority for the Financial Markets) presented the results of a questionnaire among 323 asset managers from the first half of 2025 in April 2026. Fifty-three per cent use AI or plan to within a year. A quarter have no policy on staff use of AI, more than a third have no technical or procedural controls on those tools, and 89 per cent have no policy for AI agents. At 73 per cent, all staff may freely use public generative AI tools, while 28 per cent have a policy for doing so. The AFM places final responsibility with the board and senior management and asks the sector to set down responsibilities for AI.

How solid is the evidence? Three supervisors, looking separately at banks, insurers and asset managers, find the same pattern of wide use and lagging governance, and EIOPA’s 2023 survey offers a first comparison. The picture is not settled: the bank figures come from speeches and a small sample, and the AFM figures are a snapshot from early 2025.