As artificial intelligence advances at a rapid clip, a growing number of financial institutions are beginning to use the technology, even as some warn of potential security and ethical risks.
The spectrum of opportunities that banks have in emerging AI technology was a hot topic during Febraban’s Ciab banking and technology conference in São Paulo, Brazil. Banks’ investments into AI is expected to pick up sharply in a bid to gain speed and efficiency. However, some have cautioned that banks also must put in place security frameworks to mitigate their risks.
According to Beatriz Sanz Saiz, a global analytics partner with Ernst & Young, banks need to lay the appropriate groundwork, and potentially even create new positions to address ethical concerns.
“[Banking] is the industry with the richest amount of data in the world,” she said at the Ciab conference. “Every bank needs to have a framework of ethics, transparency and accountability. Who is responsible for an algorithm? The person who collected the data? The one who developed it? Or the one who maintains it?”
Sanz Saiz noted there is still time to build a robust framework, both at the board level and through new roles for ethics responsibility. “The reality is that today algorithms are doing what we ask them to do. We are not at the point of AI being able to extrapolate learning from one area into a different one.”
“The implementation of AI will be a reputational risk that banks will need to watch,” she added.
Several presentations at the Ciab conference focused on the ethical issues and potential lines of business that might arise. AI is currently the second most invested emerging technology in the Brazilian banking sector, a study by Febraban and Deloitte showed, behind big data and analytics.
Already, some banks have started to use the technology. Guillermo Kopp, a financial services director with Microsoft, confirmed that the software giant is providing facial recognition and other services to some of Brazil’s largest financial institutions. Such technology can support customer identification, reduce the number false-positive security alerts and expedite credit score analyses.
Credit analysis and fraud management are two areas of banking in which AI already delivers concrete solutions, participants said. Many believe that’s just the tip of the iceberg.
“Today, barely 1% of devices are connected,” Laércio Albuquerque, president of Cisco Brazil, said during a conference panel. “There will be more data traffic in the coming three to five years than in all the precedent years since the invention of the internet. It will be a revolution.”
On the sidelines of the event, one expert in the field said that although banks are clearly acknowledging the benefits of the emerging technology, many still fail to see the flip side. “Many banks don’t know what they are dealing with. They don’t see the risks just yet,” the expert said.
Gustavo Fosse, chief technology officer with Banco do Brasil SA, assured that “ethics will be programmed” into its systems. “On our side, ethics is going to be not to access the data the client does not want us to. It is going to be in the hands of our consumer; it will be for him to decide what level of security he wants and how much he lets us know him,” he said.
In the long run, he said, the depth of knowledge the client allows the institution to have will affect the quality of the financial services provided, such as the interest rate charged on loans.
In August 2018, Brazil’s Congress provided a general framework on the issue when it approved the personal data protection law, a piece of regulation which seeks to mirror some aspects of the European GDPR regulation, and will come into force in 2020. “If someone shares data that he is not allowed to, there will be legislation to penalize that. It’s a crime,” Fosse said.
Fosse noted, however, a use case that still riddles bankers: What happens when a client who has agreed to share data changes his or her mind?
“How to make a machine unlearn a model is a reality that still does not have an answer,” he said.