The EU AI Act, proposed in June after two years of talks, could have implications on how financial institutions across Europe use and implement AI.   

With the passage of the act, expected this year, the EU would be the first bloc in the world to regulate AI to ensure better conditions for the development and use of the tech, according to the EU AI Act.

While no specific use of AI in financial services has been marked as a “high-risk” use case, FIs need to be “prepared for AI compliance monitoring” that comes along with the act, Timothy Spangler, partner at international law firm Dechert LLP, told Bank Automation News. 

FIs have been using AI tools for several years to target consumer marketing, identify and protect against fraud, and make credit decisions and service portfolios, Spangler said. 

“AI is too disruptive of a technology for financial institutions to simply ignore,” he said. “The impact of the act will also vary based on specific AI applications being used by a particular financial institution.” 

According to a report published by McKinsey earlier this month, the financial services industry is the “most likely to expect disruptive change from [generative] AI.” 

FIs, including SoFi and Wells Fargo, have already been leveraging AI to make lending decisions quicker and eliminating human biases. 

While the technology is able to expedite credit decisions, it can be ill-equipped to evaluate the creditworthiness of protected groups, Kareem Saleh, co-founder and chief executive of FairPlay AI, a fintech that uses machine learning (ML) models to help financial institutions reduce lending bias, previously told BAN. 

The EU is “lagging behind” in the adoption of AI compared to the U.S. and China, according to a June 2021 AI report from global analyst Forrester, but regulatory hurdles are not the main reason for this. For the report, Forrester surveyed 3,339 people in 2020 from companies with over 100 employees in Australia, Canada, China, France, Germany, India, the United Kingdom and the United States. 

According to the report, nearly 14% of all European respondents said that the inability to maintain oversight was a major hurdle for implementing the tech. Meanwhile, in the U.S., the figure was 20%. 

Regulation in the U.S.

In the U.S., the Consumer Financial Protection Bureau (CFPB) said in a report in June that the adoption of bank chatbots is expected to hit 110.9 million users by 2026. In 2022, there were 98 million users of bank chatbots, the report stated. 

The CFPB warned that as adoption increases, compliance must be top of mind for FIs. Inaccuracies from AI models and security risks need to be resolved for successful adoption of AI chatbots, the agency said. 

“Chatbots and highly scripted representatives can introduce a level of inflexibility, whereby only specific words or syntax may trigger the recognition of a dispute and begin the process of dispute resolution,” per the CFPB report. “As a result, the ability for chatbots and scripts to recognize a dispute may be limited.” 

Use of AI 

In the EU, certain uses of AI — including social scoring, real-time facial recognition and cognitive behavioral manipulation — will be banned under the AI act if passed. 

The “social scoring” of citizens can affect a person’s borrowing capability, Amit Dua, president at automated software provider SunTec Business Solutions, told BAN, noting this practice has been implemented already in China. 

“Just a simple ‘like’ button itself can throw revealing stuff,” Dua said. “Banks can stay away from some of the potential client bases where they see that there’s consistent behavior that gets demonstrated [like not being liked in their social circle].” 

Financial institutions are now beginning to assess their level of risk regarding the use of AI, Dechert’s Spangler told BAN, noting that mitigating risk could prove costly. 

“New regulations will add costs to the adoption of new AI tools,” he said. “But many financial institutions will find that the benefits of these tools outweigh the investment in expanded compliance capabilities.”