ECB study finds banks cut default risk for AI-exposed euro zone borrowers by 5%

• 2026-10-20
• News
• News Desk
• Source: European Central Bank https://www.ecb.europa.eu//pub/pdf/scpwps/ecb.wp3302~7ff8a286bc.en.pdf

Banks in four large euro-area economies lowered their assessed probability of default for borrowers in industries highly exposed to generative artificial intelligence by about 5% as public attention to AI rose after ChatGPT, according to a study published by the European Central Bank. The research, based on lender-specific default-risk estimates for private firms, describes banks as interpreters of the technology's spread across companies.

Key facts

• The study covers private firms in four large euro-area economies and uses banks' own probability-of-default (PD) estimates.
• Banks cut assessed default risk for borrowers in highly AI-exposed industries by about 5% after attention to AI rose following ChatGPT.
• The favorable industry-level effect was weaker for labor-intensive firms and stronger among lenders with pre-existing industry expertise.
• Within product markets, banks reassessed firms with stronger pre-ChatGPT market positions less favorably than competitors.
• Market leaders saw relatively lower drawn and committed credit, more relationship exits and fewer new lenders.

The study argues that general-purpose technologies create large opportunities but make it hard to judge in real time how their benefits will be distributed across firms. To get around that problem, the researchers treated banks as interpreters of generative AI's rise, examining how lenders' default-risk assessments for private companies shifted as the technology gained public prominence.

Why did banks lower risk for AI-exposed firms?

The findings suggest that lenders read strong exposure to generative AI as a positive signal for borrowers' prospects. Once public attention to AI climbed after the launch of ChatGPT, banks marked down the assessed default risk for companies in highly AI-exposed industries by roughly 5%.

That reassessment was not uniform. It was weaker for labor-intensive firms, which the study links to the different way such companies are expected to absorb the technology. It was stronger among lenders that already had expertise in the relevant industries, indicating that familiarity with a sector shaped how banks read the AI signal.

What happened to market leaders?

The favorable industry-level effect came alongside a systematic reordering within product markets. Banks reassessed firms that had stronger market positions before ChatGPT less favorably relative to their competitors, compressing the risk advantage that market leadership had previously conferred.

Credit allocation moved in the same direction. Market leaders experienced relatively lower drawn and committed credit, more exits from banking relationships and fewer new lenders, according to the study. The reordering also predicted subsequent bad-credit events beyond the banks' consensus default-risk estimates, the research found.

The authors conclude that internal bank assessments offer a forward-looking measure of how generative AI is expected to reshape competitive risk, rather than a simple snapshot of current conditions. For European lenders, the results point to a channel through which the technology could redistribute credit even before its effects show up in conventional default statistics.

The study does not specify a date for publication of further findings or a follow-up assessment, and it offers no timetable for when the shifts in credit allocation might feed through to broader lending patterns in the euro area.

Source: European Central Bank