AI Act and the Banking Sector: The Dossier on the Brussels-BBVA Axis
- BBVA was the first bank to meet officially with the European Commission to define the application of the AI Act in the credit sector.
- The focus of the talks concerns the definition of AI system and the classification of high-risk systems to ensure consistency across Member States.
- The EU is implementing a new supervision team to monitor large AI models, including OpenAI and DeepSeek.
- There is an open debate between the urgency of implementing the regulation and proposals to postpone some key points by over a year.
The integration of artificial intelligence into financial services is no longer just a technological challenge, but a matter of legal and strategic compliance. The recent rapprochement between Brussels institutions and major banking players marks the beginning of a crucial operational phase for the AI Act, with direct implications for risk management and credit operations throughout the European Union.
The secret June meeting: why BBVA became Brussels' first banking interlocutor
On June 9, at the central offices of the department responsible for digital policy in Brussels, a meeting took place that set an institutional precedent. As reported by Economía Digital, the European Commission (EC) convened representatives from BBVA to study the impact and application of the Artificial Intelligence Act specifically in the banking sector.
The importance of this meeting lies in the fact that it was the first official meeting between the European Union and a single banking institution to address this matter. The main interlocutor on behalf of the EC was Thierry Boulangé, interim head of the unit of the Directorate-General for Communications Networks, Content and Technology. The Commission's goal was to understand how the rules of the game, which are changing globally, interface with the operations of an international-scale banking group.
Strategic Analysis: The choice of BBVA as the first interlocutor is not accidental. The Spanish group has recently accelerated its technological strategy, creating a new area dedicated to AI led by Antonio Bravo. For Brussels, collaborating with an entity that has already internalized AI at a structural level allows for testing the technical feasibility of the rules before their generalized application.
Legal concepts at stake: the battle over the definition of AI and application among Member States
The dialogue between BBVA and the European Commission was not a simple exchange of courtesies, but a technical table focused on critical legal nodes. Below is the analysis of the points discussed in a question and answer format, based on documentation from the EC transparency portal.
What is the main objective of the definition of an AI system?
The priority is to establish a 'definition of the artificial intelligence system and its consistent application across all Member States'. The goal is to prevent a bank operating in multiple EU countries from having to comply with divergent interpretations of the rule depending on the national jurisdiction.
What are the legal concepts that concern the sector?
The working group focused on the definition of certain legal concepts that could influence banking operations. Although specific details have not been made public, attention is focused on how the AI Act overlaps with pre-existing banking regulations in terms of liability and transparency.
How does this consistency affect competitiveness?
A unique definition allows credit institutions to scale their AI solutions across the entire Schengen area without having to redesign the compliance architecture for every single market.
The high-risk perimeter: how guidelines for the credit sector are drafted
One of the most delicate points of the AI Act is the classification of AI systems based on risk. For the banking sector, the boundary between 'limited risk' and 'high risk' determines the extent of bureaucratic and technical burdens.
Behind the scenes of the June meeting, participants worked on a borrador de directrices (draft guidelines) on high-risk classification. In the credit context, high risk is typically associated with systems that influence access to essential services, such as:
- Credit Scoring: Algorithms that decide the granting of a loan or the reliability of a customer.
- Risk Assessment: Automatic monitoring systems for anti-money laundering prevention or credit risk management.
- Decision Automation: Processes that replace human intervention in decisions that have significant legal effects on the user.
Business Analysis: If a scoring system is classified as 'high risk', the bank will have to implement rigorous data governance systems, detailed technical documentation, and constant human supervision. This increases implementation costs but reduces the risk of million-euro sanctions provided for by the AI Act.
From Spanish bankers to the German association: the extension of the EU-Finance dialogue
Although the meeting with BBVA was the first with a single institution, the European Commission is adopting a multi-level approach to map the needs of the financial sector. The map of the actors involved shows a diversified consultation strategy:
| Actor | Type of Interlocutor | Timing | Dialogue Objective |
|---|---|---|---|
| Bundesverband deutscher Banken | Trade association (Germany) | January | Representation of the collective interests of German banks. |
| BBVA | Single Banking Institution (Spain) | June | Study of the practical and technical application of the AI Act in a global group. |
| European Commission (DG Connect) | EU Regulator | Continuous | Drafting of guidelines and supervision of implementation. |
This shift from a dialogue with associations (such as the German one) to meetings with single players (such as BBVA) indicates that Brussels has moved from the political listening phase to the technical validation phase.
Active supervision: how the new control team for OpenAI and DeepSeek will work
Parallel to the dialogue with banks, the European Union is tightening control over providers of general-purpose AI models. As reported by El Imparcial, Brussels has put in place a new supervision system with the entry into force of the AI Act.
The new supervision team will be tasked with monitoring companies such as OpenAI and DeepSeek. The control strategy is based on three pillars:
- Transparency Monitoring: Verifying that models are trained on data that respects copyright and EU rules.
- Systemic Risk Assessment: Analysis of the models' capacity to cause large-scale harm or influence democratic processes.
- Technical Audits: Possibility of requesting access to training data and optimization processes.
Geopolitical Analysis: The inclusion of DeepSeek (a Chinese company) on the EU supervision radar underlines Brussels' desire not to limit control only to US players, but to apply the European standard to any technology that accesses the single market, regardless of geographic origin.
The timing dilemma: between the urgency of implementation and the proposal for a one-year postponement
Despite the push toward supervision, there is internal tension between the need to regulate quickly and the technical reality of implementation. According to Expansión, a proposal has emerged from Brussels to delay by over a year some key points of the Artificial Intelligence Act.
The comparison between the two positions highlights the critical issues of the moment:
- Pro Postponement (Prudential Approach):
- Allows companies to adapt infrastructure without interrupting services.
- Avoids the application of rules based on definitions still in draft phase (such as those discussed with BBVA).
- Reduces the risk of 'over-regulation' that could stifle European innovation compared to the USA and China.
- Against Postponement (Urgency Approach):
- AI evolves at such a speed that a one-year delay could make the rules obsolete before they even enter into force.
- Lack of immediate protection for citizens against high-risk AI systems already in operation.
- Loss of EU regulatory leadership in defining the global standard (the so-called 'Brussels Effect').
Future Scenario: If the postponement is approved, we will witness a period of regulatory 'grey zone'.
Verifiable Indicator: The publication of the updated calendar of AI Act compliance deadlines by the European Commission within the next semester.
Banking AI and European compliance: the impact of the AI Act on risk management and the operations of Italian banks
For entrepreneurs and executives in the financial sector in Italy, the Brussels-BBVA axis is a clear signal: compliance will not be an automatic process, but the result of technical negotiations between the regulator and the industry.
The impact of the AI Act will manifest on three operational levels:
1. Risk Management and Governance: Italian banks will have to map every single algorithm used. If a creditworthiness assessment system falls within the high-risk perimeter, the institution must prepare technical documentation that justifies every automated decision, making it explainable (explainability) to the end user.
2. Integration with NIS2: The AI Act does not operate in a vacuum. Its application intertwines with the NIS2 directive on cybersecurity. Banking AI systems must not only be 'ethical' and 'transparent', but extremely resilient to cyber attacks, as a compromised AI in a critical sector like banking represents a systemic risk for the entire economy.
3. Local Market and Competitiveness: Small and medium-sized Italian banks could find themselves at a disadvantage compared to large groups (like BBVA) that have the resources to participate in Brussels' tables and influence guidelines. This could accelerate banking consolidation processes, where the ability to manage technological compliance becomes a strategic asset.
'The definition of the artificial intelligence system and its consistent application across all Member States' is the pivot upon which the ability of European companies to compete without being stifled by bureaucracy will play out.
In conclusion, the EU is seeking a precarious balance between the protection of fundamental rights and the need not to fall behind in the technological race. For Italian companies, the key will be anticipation: not waiting for the final publication of the guidelines, but beginning to classify their own AI systems according to the risk criteria discussed between the Commission and the first banking interlocutors.
Summary for the entrepreneur: The AI Act transforms AI from a competitive advantage into a compliance obligation. Those who invest today in 'governed' and transparent AI will have a competitive advantage tomorrow, avoiding costly forced redesign processes.
FAQ
Why is the meeting between BBVA and the European Commission considered significant?
It was the first official meeting between the EU and a single banking institution to define the practical application of the AI Act in the credit sector, shifting the dialogue from political theory to technical validation.
What is meant by 'high-risk' AI systems in banking?
These are systems that influence access to essential services, such as credit scoring algorithms for loan granting or risk assessment systems that have significant legal impacts on customers.
Which companies are under the supervision of the new EU control team?
The new supervision system will monitor providers of general-purpose AI models, explicitly citing companies such as OpenAI and DeepSeek.
Is it possible that the application of the AI Act be postponed?
Yes, a proposal has been put forward by Brussels to delay the implementation of some key points of the law by over a year to allow for a more gradual adaptation by companies.
Sources: Economiadigital, Expansion, Elimparcial · by glacom.news AI
Scrivila qui: Susanna, l assistente AI di glacom, ti risponde via email con un approfondimento gratuito.
Nessuna consulenza personalizzata (finanziaria, legale o medica): solo informazione e fonti. Email usata solo per rispondere.
.png)
