intervenção pública01/02/2024Intervenção de abertura do Administrador Hélder Rosalino na Fintech Meeting: Credit risk models | Machine Learning (apenas em inglês)

Good afternoon!

Thank you all for joining us for another Fintech Meeting of the Banco de Portugal, the first one of 2024.

And thank to our guest speakers, for being here today with us.

It is a pleasure to have you at this meeting, with a very significant industry representation.

The topic that brings us here today is the use of artificial intelligence in the banking sector, in general, and applied to credit scoring, in particular.

This topic is of the utmost importance, as we will see.

Before passing on the word to our distinguished speakers, I would like to convey some thoughts on the relevance of this topic for financial supervisory authorities, spreading across most of our missions, namely in the areas of financial stability and consumer protection.

The term artificial intelligence is not new. It was coined in 1955[1] and has been around ever since.

However, it has only recently caught the attention of a more comprehensive community, fueled by its technology's commercial success (pe: using ChatGPT and other tools like that, generally exploited by all in daily basis) and the by increasing the fields of its application in several economic sectors.

This recent uptick has been based mainly on three factors[2] :

• the availability of data (nowadays, more than ever);

• storage capacity and computation power, which in turn were propelled by complementary technologies; and

• other digital developments like social media, electronic commerce, cloud computing and graphic processing units.

Because of these reasons, we have witnessed a shift in the way "production factors" are allocated and remunerated in almost all economic sectors due to the impact of artificial intelligence techniques in the daily production, storage, distribution, marketing and sale of products and services from firms to customers[3]. In fact, we can see a major transformation in the economy as a whole by the use of AI.

Particularly in the financial sector, the level of adoption of artificial intelligence by banks and banking-related entities is now relatively high.

This can be seen as part of a broader process of digital transformation, stemming from the need to adapt to changing customer preferences for digital based, tailor-made, quick solutions, and growing competition from new players in the market and between the incumbent players.

In this regard, the last information published by the European Central Bank in the Supervision Newsletter[4] noted that banks still face some deficiencies in their digital transformation strategies. However, with significant adoption of AI solutions, at least on an experimental basis.

Obviously, it is essential to separate the adoption by banks and other market players from the adoption by Central Banks and other supervisory authorities.

From a supervisory perspective and in tune with the European Central Bank, Banco de Portugal has been exploring the opportunities in several applications and the risks they entail in terms of data privacy, legal constraints, and ethical considerations, including fairness, transparency and accountability[5].

In this regard, I would like to mention that Banco de Portugal recently created ALYA, an AI platform with machine learning and deep learning capabilities, which has allowed the automation and efficiency of activities in various domains of our core functions.

And we are happy to be able to say that ALYA received the Portugal Digital Award 2023 for the best digital transformation project in the public sector.

Regarding the adoption by banks and other market participants and building upon the report recently published by the European Banking Authority[6], we are aware of the current use of AI algorithms in risk management, customer relationship management, fraud detection and back-office automation.

A more visible part of this AI shift is the famous "Chatbots" that alleviate the burden on customer support teams by providing inferred answers.

Fraud detection and back-office automation have mainly upgraded in speed, effectiveness, and efficiency the banks' operations and benefitted greatly from the new algorithmic approaches.

Another relevant field is risk management, which involves several pertinent areas, such as credit risk, which is today's focus.

This topic is of enormous importance for Banco de Portugal because of its significant financial stability and consumer protection implications.

Credit scoring is classified as a "high-risk" application under the European Commission's new proposal for AI regulation[7].

Although the final text has not yet been approved, and we could yet see some changes, namely considering the opinion of the European Central Bank regarding its tasks under that Regulation, we believe it is safe to say that supervisory authorities will be fundamental in its implementation and must, therefore, be prepared for this.

Here at the Banco de Portugal, we follow closely and generally agree with the opportunities and concerns surrounding using artificial intelligence for credit scoring.

On the one hand, AI for credit scoring could give a greater ability to process and analyse information may improve credit risk forecasts, allowing for a more effective and robust determination of the risk-remuneration binomial.

On the other hand (on the danger side), and among the usual voiced concerns, we highlight discrimination and credit concentration risks as the ones with the most potential impact on consumer protection and financial stability.

And this is why we have organized this FinTech meeting now.

Until now, we see a relatively low level of use, risk, and concern with the issue, but we want to bring the topic to everybody's radar so that we can prepare on time for what is coming.

The future of the lending business[8] will most likely be based on these disruptive technologies. It is up to us to understand, assess and control the risks stemming from that transformation.

For that, we think it is crucial to have an ecosystem that is working on this topic, close to supervisory authorities, with the necessary funding, a clear regulatory framework and a common mindset of concerns.

To conclude, I am sure that we will have a fruitful discussion and I hope that you all enjoy the session.

I will now pass on the word to the European Commission.

Thank you.

[1] Artificial Intelligence (AI) Coined at Dartmouth | Dartmouth

[2] (PDF) A Brief History of AI: How to Prevent Another Winter (A Critical Review) (researchgate.net)

[3] Artificial Intelligence in Society | en | OECD

[4] Banks' digital transformation: where do we stand? (europa.eu)

[5] Careful embrace: AI and the ECB (europa.eu)

[6] EBA report on Big Data and Advanced Analytics | European Banking Authority (europa.eu)

[7] EUR-Lex - 52021PC0206 - EN - EUR-Lex (europa.eu)

[8] building-the-ai-bank-of-the-future.pdf (mckinsey.com)

Public link

Please use the above public link if you want to share this noodl on another website.