FT Longitude found 63% of banking executives would consider less transparent artificial intelligence (AI) for better performance, while 18% said their banks had fully operational, embedded governance processes for AI models.
FT Longitude surveyed 300 banking executives in 31 countries in June and July 2026; SAS and Microsoft sponsored the research. Ten percent of respondents said AI already made autonomous credit or risk decisions at scale in their banks. Seventy-six percent expect it to do so within five years, with 25% expecting it within two years.
In the survey, respondents linked their willingness to accept limited transparency to a model offering better performance, return or liquidity than an explainable option.
Other oversight measures were less common. Nineteen percent said their banks monitored model drift in real time, and 36% reported continuously retraining or updating models. Forty-two percent said their institutions lacked “kill switches” to restrict damage when a model crossed preset boundaries.
The report found that 92% of respondents said at least one AI model at their institution ran outside formal governance frameworks. On average, 20% of models were outside those frameworks, according to the report.
Executives also cited barriers to giving AI more decision-making authority. Seventy-nine percent named a lack of trust in data quality as a main or contributing reason for limiting autonomous decisions, while 77% cited regulatory or compliance concerns.
The report linked willingness to use less-governed models partly to fear of missing AI opportunities. It said AI could benefit banks and their clients, but warned that black-box models could harm communities and cause financial or reputational damage to banks.
Marilin Pikaro, the European Banking Authority’s director of innovation, conduct and consumers, said in the report: “Ambition should match its capacity.” In its June 2026 risk assessment, the European Banking Authority (EBA) said AI risks, if they materialize, could cause financial losses, reputational damage or systemic consequences.
