What does EBA Risk Assessment Report (Nov 2026) say about AI in banking?

Published Nov 19, 2026 · Last reviewed Nov 19, 2026

The EBA's Autumn 2026 Risk Assessment Report, published on 19 November 2026, says EU/EEA banks expect AI to improve productivity and are already making extensive use of frontier AI models, but that adoption adds model risk, dependence on external providers and exposure to cyberattacks and fraud. Its Risk Assessment Questionnaire found that nearly three quarters of the banks covered have access to frontier AI models, nearly two thirds already use AI models to detect or remediate software flaws, and around 80% source the AI models they use from the US. The report creates no requirements, but it lists AI risks it says banks need to identify and effectively manage.

OFFICIAL TEXT: eba.europa.eu ↗ · FINAL · EBA

Checked 2026-11-19 · EBA, eba.europa.eu
DocumentEBA Risk Assessment Report (Nov 2026) — Risk Assessment Report of the European Banking Authority, November 2026, with Box 1 on artificial intelligence related risks and opportunities
Issued byEuropean Banking Authority
TypeReport
StatusFinal
PublishedNov 19, 2026
Applies toThe EU/EEA banking sector; findings draw on supervisory data and the EBA's Risk Assessment Questionnaire (RAQ) of banks. A risk assessment, not guidance
Official sourceeba.europa.eu ↗
Use casesCybersecurity · Generative & agentic AI · Third-party & vendor AI · Fraud detection · Model risk management · AI governance (general)

What are the key points of EBA Risk Assessment Report (Nov 2026)?

  • Box 1 covers three channels: AI adoption in banks' business and operations, AI-related cyber and ICT risk, and exposure to the financing of AI-related investment.
  • Nearly two thirds of banks in the RAQ expect AI to reduce operating expenditure, mainly through employee productivity; one bank expects a negative cost impact, and smaller banks are more cautious about AI's overall effect.
  • Risks named from AI use: inaccurate or unreliable outputs in critical processes, loss of human control and faster shock propagation when tasks are delegated to AI agents, inability to understand and validate third-party models, automation bias, and concentration on AI providers, many headquartered in the US.
  • Nearly all banks say recent cyber-threat developments, including AI, changed their approach to cybersecurity (about 30% significantly, nearly 70% to some extent), most often in vulnerability management and monitoring.
  • The EBA warns that EU banks may not be able to access frontier AI models for defensive purposes because most developers of such models are headquartered outside the EU.
  • Around 80% of banks source AI models from the US and around half that number from European suppliers; Chinese and other models are only marginally used.
  • AI could raise funding costs indirectly if clients use AI tools to compare returns and move money into higher-yielding deposits or investments.
  • The financing needs of the AI ecosystem warrant monitoring of banks' related concentration and valuation risks, although banks have on average only limited exposures to the ICT sector.

What did EBA Risk Assessment Report (Nov 2026) change for banks?

Earlier EBA AI publications tracked adoption (the September 2025 report on AI in EU banking and payments) and mapped the AI Act onto banking rules. This report puts frontier AI into the EBA's assessment of risks to the whole EU/EEA banking sector, with questionnaire data on access, defensive use and supplier geography, and it cites the ESAs' July 2026 statement on ICT risks from frontier AI models (JC 2026 25) as the supervisory reference.

What does the EBA's November 2026 Risk Assessment Report say about AI?

That EU/EEA banks expect AI to raise productivity and already use frontier AI models, including to find and fix software flaws, but that AI can increase exposure to cyberattacks and fraud, add model risk and create dependence on external, mostly US, providers. Box 1 also flags loss of human control when tasks are delegated to AI agents, automation bias and possible pressure on funding costs.

How many EU banks have access to frontier AI models, according to the EBA?

The EBA's Autumn 2026 Risk Assessment Questionnaire found that nearly three quarters of the banks covered have access to frontier AI models and that nearly two thirds already use AI models to detect or remediate software flaws and system weaknesses. The EBA notes that smaller banks outside its sample may be more cautious than those it surveyed.

Is the EBA Risk Assessment Report binding?

No. It is the EBA's assessment of risks and vulnerabilities in the EU/EEA banking sector, based on supervisory data and a questionnaire of banks. It creates no new requirements, but it says banks need to identify and effectively manage the AI risks it describes, and it refers to the ESAs' July 2026 statement on ICT risks from frontier AI models.

DateDocumentStatus
Nov 5, 2026EBA/GL/2026/10 (internal governance) — EBA/GL/2026/10 Guidelines on internal governance under Directive 2013/36/EUFinal
Sep 18, 2026EBA/GL/2026/09 (third-party risk) — EBA/GL/2026/09 Guidelines on the sound management of third-party risk related to non-ICT servicesFinal
Jul 31, 2026ESA Statement on ICT risks from frontier AI models (JC 2026 25) — ESA Statement: Toward a consistent and risk-based approach for ICT risks from frontier AI modelsIn force
Nov 21, 2025EBA factsheet: AI Act implications for the EU banking and payments sector — AI Act: implications for the EU banking and payments sectorFinal
Nov 21, 2025EBA Chair letter to the Commission on the AI Act mapping exercise (EBA/2025/D/5384) — Outcome of EBA's AI Act mapping exercise — letter to DG FISMA and DG CNECTFinal
Oct 1, 2025EBA Work Programme 2026 — EBA Work Programme 2026 — AI Act implementation and digital-finance prioritiesIn force

Follow every document these regulators publish

when one of these regulators moves, the next morning's brief says so · six sourced stories · 7 am PT · free

plus every regulation, bank and agent page update, the morning after · leave any morning