About half of all staff in Luxembourg’s financial services sector is permitted to use generative artificial intelligence tools on their office networks, while a quarter - 24 per cent - of all firms have blocked access to tools such as ChatGPT, according to a new survey presented on Thursday by banking association ABBL.
At banks, the usage of ChatGPT and other generative AI tools is even more restricted, with 38 per cent of staff being blocked from accessing it, the survey showed. In this regard, ABBL is asking for more clarity from data protection and financial services authorities. This “would be useful on the usage of generative AI in the banking sector,” said ABBL when presenting the survey. “This will be an important signal to credit institutions to proceed with more enhanced experimentation with the technology.”
“We observe that the adoption of the generative AI technology in the banking sector is held in a more regulated context. This could be reflected in a relatively higher level of internal restrictions and hence could lead to the limited scope and pace of adoption,” said Laurent Marochini, chair of ABBL’s FinTech and Innovation Forum and head of innovation at Société Générale Securities Services Luxembourg, in a statement.
Luxembourg’s financial supervisor CSSF, in close collaboration with the country’s central bank BCL, recently presented its own report on the use of ChatGPT and other generative AI tools. It noted that use of AI at this time still is fairly limited and still at an early stage, but that it expected to encounter wider adoption in the future.
Seen as new era
The ABBL survey was conducted among 42 respondents representing credit institutions, FinTech firms, consulting companies, payment institutions, electronic money institutions, support firms, law firms, and software vendors, and was supported by Sociéte Générale. It showed that for 51 per cent of them, ChatGPT is free to be used by all staff members. Partial use by certain categories of staff is permitted by 24 per cent.
ABBL said the survey showed that three quarters of the respondents are optimistic about the benefits and opportunities offered by ChatGPT.
“The release of ChatGPT marked the beginning of a new era for business,” said Marochini. “Language model implementation has already passed the point of no return. It will impact the functioning of society through its implementation in business processes. Although changes will not be as revolutionary for all, some will be affected, and the first movers will be rewarded with lower costs, increased speed, and greater accuracy.”
The survey showed that the majority of respondents – 57 per cent – have planned to implement ChatGPT or other AI-powered chatbots for specific use cases. This figure becomes more striking for credit institutions: 88 per cent. “This is an indication of a strong credit of trust given by banks towards the Generative AI technology, and we expect that the adoption journey is likely to accelerate in the coming years”, pointed out Andrey Martovoy, innovation & digital adviser at the ABBL.
Positive impact seen, also on bank profits
The impact of artificial intelligence is widely expected to reduce the need for basic task-performing labor, leaving more time to focus on creative and complex tasks. “ChatGPT and the like will have a positive impact on efficiency and could be part of the answer to growing pressure on banks’ profitability,” said Ananda Kautz, head of innovation, digital banking & payments at ABBL.
More personalized banking services are expected to emerge as a possible result of generative AI.
“Respondents of the survey highlight opportunities offered to improve customer services. Examples could include more personalized banking experiences thanks to enhanced customer analytics, investment advice based on larger and new sources of data or customer service chatbots available 24/7,” said Kautz. “Finally, ChatGPT can support financial institutions in detecting fraud by analysing transactional data in real-time, and provide support to the Tech team. In a context of scarce Compliance and IT resources, ChatGPT could be part of the solution.”
Hallucination
ABBL’s innovation experts also highlighted some of the dangers stemming from generative AI tools. Inaccuracy, also described as “hallucination”, is seen among the most threatening aspects.
“Language models can sometimes completely invent the answer to a question phrased by the user,” said Kautz. “This phenomenon can be difficult to identify due to the persuasive arguments used. The only way to counter it is to systematically check the sources on which the models base their argumentation. It implies that one cannot completely rely on this technology for learning purposes.”
Data privacy is another issue of concern. Every input sent into a language model goes through the servers of its creator: Open AI in the case of ChatGPT. This means the model cannot be used to process confidential data, especially ones protected by GDPR.
Nevertheless, AI tools are widely expected to bring efficiency gains. ABBL said commercial entities are inherently incentivised to continue developing and implementing solutions by making them compliant to existing regulatory frameworks and standards. “The results show that 76% of respondents consider tools such as ChatGPT an opportunity for their organizations,” said Kautz.