Bitget: haz trading inteligente. Lionel Messi, Official Partner. Regístrate

Its risk report sees European banks as solid, with top-quality capital (CET1) of 16.2% in June. Banks are already using AI, including frontier models, to find weaknesses in their systems, but the EBA warns that it may increase exposure to cyberattacks and fraud, add model risk and create dependence on third-party providers. It calls for monitoring concentration and valuation in financing linked to the AI ecosystem

European banks are reaching the end of 2026 with good marks. The European Banking Authority (EBA) published its autumn 2026 risk assessment report on November 19, and the overall diagnosis is that banks in the European Union and the European Economic Area have more than enough capital and liquidity to absorb a shock. The fine print is less reassuring. The supervisor points to three fronts worth watching, and one of them, technology, is of direct interest to anyone following the blockchain world.

The figures behind European banks’ passing grade

The top-quality capital ratio (CET1), which measures a bank’s own cushion for absorbing losses, stood at 16.2% in June 2026. That leaves a margin of 4.7 points above what supervisors require. To put it simply, it is like setting off on a trip with a tank much fuller than the journey needs.

Profitability is keeping pace. Return on equity (RoE) was 11.3% in the second quarter, compared with 10.7% a year earlier. Liquidity is also comfortable, with a liquidity coverage ratio (LCR) of 158.5% against a legal minimum of 100%. And non-performing loans remain low, at 1.83% of lending.

In addition, banks are lending more. Lending to businesses rose 6.4% year on year and lending to households 5.3%.

Public debt weighs increasingly on balance sheets

European banks hold around €4.5 trillion in government debt. That figure equals 244% of their CET1 capital, up from 206% in June 2024. Sovereign bonds also now make up 47% of the liquid asset buffer, more than double the 22% of June 2022.

The problem arises when the yield the market demands on that debt rises, because the bonds the bank already holds lose value. About 40% of the exposure is accounted for at market value, leaving around €1.6 trillion exposed to valuation losses. The EBA estimates that a 100 basis point rise (one percentage point) would cost more than half of institutions less than 2% of their Tier 1 capital. It is a manageable hit, but the health of the banking sector is more tied to that of public finances than it was two years ago.

SMEs, the weakest link in lending

The average non-performing loan ratio is low, but the aggregate figure masks what is happening underneath. The volume of non-performing loans rose 4% in one year, to around €388 billion, and new defaults grew by about 8% year on year in the first half. The ratio barely moves only because total lending is growing at a similar pace.

The focus is on small and medium-sized enterprises. Their non-performing loan ratio is 4.4%, more than double the average, and 14.2% of their loans are in the watch category (what accounting standards call stage 2, loans that are still being repaid but whose risk has increased). The EBA devotes a specific note on SME lending to the issue and warns that a significant share of those loans overlaps with commercial real estate exposure. Put plainly, a stumble in offices and retail premises would hit the same portfolio twice.

Another figure stands out. Lending to non-bank financial institutions, such as investment funds, grew 18.5% year on year and already totals €1.74 trillion. It is the channel through which a problem born outside the banking sector can end up inside it.

AI and quantum computing, the risk that also affects blockchain

Here is the heart of the matter for Territorio Blockchain readers. The EBA places operational risks (cyberattacks, fraud and dependence on technology providers) among the main threats, and artificial intelligence plays for both teams. Almost two thirds of the banks surveyed already use it to detect or fix flaws in their software, and almost three quarters have access to the most advanced models. But around 80% obtain those models from US providers. It is a dependence reminiscent of the dollar’s dominance in stablecoins, and it ties in with the debate on how to oversee the most advanced AI within banks that the US supervisor has already opened.

The second issue is quantum computing. A sufficiently powerful quantum computer could break the public-key cryptography that currently protects banking communications, the same family of mathematics that protects Bitcoin or Ethereum signatures. The report recalls the “harvest now, decrypt later” tactic, which consists of stealing encrypted data now and storing it until a machine capable of unlocking it exists.

The readiness numbers are weak. 66% of banks aim to migrate their highest-risk use cases to post-quantum cryptography by 2030, but only 8% have a roadmap underway. Just 34% have a complete and up-to-date inventory of the cryptography they use, and 73% have not documented what plans their providers are following.

What it means for you

If you are a bank customer, the immediate message is reassuring. There is capital, there is liquidity and non-performing loans are contained. What is changing is the type of risk. After the 2008 crisis, the fear was that a bank had lent badly. Today the EBA is also looking at a technology provider that fails, an AI model that makes mistakes or an encryption system that becomes obsolete.

European banks’ good grade is real, but the exam of the coming years increasingly resembles the one the crypto sector has long been preparing for. Who do you think will reach post-quantum cryptography first, the banks or the blockchain networks?

Get the best of the week in your inbox

Every Monday: the most-read stories on Territorio Blockchain, the new podcast episode and news about our events. Free, no spam.