Compliance experts ID 3 must-dos when using AI in auto lending

As regulatory scrutiny intensifies across the auto finance industry and AI, lenders should know where AI exists in their processes, outline AI governance and watch vendor partners. Auto lenders that use AI may face regulatory “double jeopardy,” as state regulators ramp up scrutiny of the industry and federal regulatory bodies create AI guidelines, Mark Metrey , associate focusing on AI, privacy and credit reporting laws in litigation and government enforcement actions at law firm Hudson Cook , said during a session at the recent Non-Prime Auto Financing Conference. “The tough thing with AI is … it’s being embedded in everything, meaning there is an even higher level of scrutiny.” — Mark Metrey, Hudson Cook “There is so many different aspects of the auto industry that already have been for a long time heavily scrutinized, and the tough thing with AI is … it’s being embedded in everything, meaning there is an even higher level of scrutiny in all those same areas,” Metrey said. For example, Nissan Motor Acceptance Co . paid $5.5 million in penalties and restitution and will provide refunds to New Yorkers who were overcharged when buying out leased vehicles, according to a June 3 settlement announced by state Attorney General Letitia James . President Donald Trump issued an executive order June 2 aiming to strengthen cybersecurity by requiring AI companies to voluntarily provide the federal government access to new models for up to 30 days before their planned release. Auto lenders should make three considerations when using AI to minimize risk, Metrey said. 1. Know where AI exists Lenders need to be able to explain where AI is used in underwriting and how it makes decisions or they could incur major penalties, Metrey said, referencing student loan company Earnest Operations ’ $2.5 million settlement with the Massachusetts attorney general’s office in July 2025. Earnest settled with Attorney General Andrea Joy Campbell after she alleged the company’s use of AI underwriting models included a default rate based on what schools students attended, resulting in disparate penalization of Black and Hispanic applicants, according to a release. The company did not admit to any wrongdoing in the settlement. The company was unable to explain how its AI came to decisions, which signals “you don’t know the tool that you are using, and that creates an immense amount of risk,” Metrey said. Lenders using alternative data beyond traditional FICO scores in AI underwriting models could be at risk of making biased decisions if the AI is not accurate or the criteria is subjective, he added. 2. Build governance and accountability Once lenders have identified where they are using AI, they should categorize the risk of those AI tools, Hudson Cook Partner Megan Nicholls said during the panel. Nicholls focuses on consumer information, fraud and data security. To determine risk, she said, lenders should know whether the AI is developed internally and whether its agentic, which could make decisions on its own. “The next step is to start categorizing that risk for each of those particular models, and the risk classification goes back to your overall risk governance structure,” she said. Lenders should build out their AI governance structure, according to the presentation, by: Assigning ownership of AI processes; Defining approved and prohibited AI uses for employees; and Requiring documentation and explainability standards. Once risk is outlined and assigned, lenders should regularly check the AI model to ensure it is not hallucinating in ways that incur risk, Metrey said. “It might be completely inadvertent, but if there is a pattern of potential bias in the [AI’s] outputs, that can create a lot of issues,” he said. 3. Strengthen vendor oversight When working with third-party AI vendors, lenders should outline which entity is liable for actions the AI takes, especially in cases of “hallucinations or bad outputs,” Metrey said. “Do not assume that whatever the vendor guarantees, even in the contract, is automatically federal and state complaint. — Mark Metrey, Hudson Cook Lenders should also complete data security reviews with vendors, making sure to clarify whether the data used will be owned by the AI vendor and used across all the vendors’ models, or if it’s exclusive to the lender, Nicholls added. “There are two additional questions that we’ve stumbled on … How’s [AI] being used and how is it being developed?” Nicholls said. Auto Finance Summit, the premier industry event for auto lending and leasing, returns October 5-7 at Caesars Las Vegas. To learn more about the 2026 event and register for early-bird pricing through Aug. 21, visit www.AutoFinance.live/AFS .

Compliance experts ID 3 must-dos when using AI in auto lending

Knowing how AI underwrites can prevent millions in settlements

Exclusive: Brex, Tekion launch dealer-facing expense management platform

Related Posts

NMAC issuing lease buyout refunds following settlement with New York AG

Santander Consumer USA agrees to $400K settlement with New York

FTC names 97 dealership groups that received pricing warning letters

Potential disclosure regulation reform could heighten enforcement risk for ABS issuers

Stay Informed with Our Newsletters

The Roadmap Podcast

SPONSORED

Why credit unions give dealers an edge in today’s auto market

Driving better decision-making across auto finance operations with SAS

Auto finance’s first line of defense: Raising the standard in integrated software partnerships and data strategy

Retrieve your password