Moody’s warns of risks in AI taking central role in lending
The use of artificial intelligence in loan origination is rising, with some structured finance sponsors utilising the technology to source assets, according to a new Moody’s report.
Although financial institutions have been incorporating AI into their lending processes for years, some are now explicitly using AI models to help make underwriting decisions and select loans. The European Banking Authority recently revealed that only 26 per cent of responding institutions to a recent survey were not using or planning to use AI in credit scoring.
Examples of groups that have put AI at the centre of their proposition include lending platform Upstart, auto loan originator Lendbuzz and consumer loan provider Pagaya.
Read more: Artificial intelligence backed to reduce P2P defaults
While AI can help sponsors use alternative data to improve their accuracy, Moody’s also warned that it comes with risks.
“We generally expect that pools with assets underwritten by AI-based models will have relatively volatile losses, in part because of the short performance histories. In general, the shorter a lender’s history, the less predictable their loan performance is, especially as economic conditions change,” the report noted.
“In addition, AI models have not been tested through many economic cycles, and could perform differently than expected during future downturns. Their accuracy could also worsen if they are extended to additional products or geographies, depending on the breadth of data over which the models are trained, because borrowers’ behavior may differ by the type of loan product or the region where they live.”
There are also regulatory risks, Moody’s warned, because in most jurisdictions the use of AI in such cases is heavily regulated. Noncompliance when using AI in lending could lead to fines, penalties and reputational damage for the sponsor.
Read more: How AI could change the future of P2P
Read more: Zopa signs up to artificial intelligence credit scorer Aire
