Q

Conference & Expo: TBD, 2027
DealerPoint: April 5-7, 2027

Q

Auto Lenders Face a Patchwork of AI Rules and Governance Changes from Mid-term Election Outcomes

Published: November 20, 2026

Artificial intelligence now supports underwriting, fraud screening, collections and customer service across auto lending. The rules governing those tools have not settled, and the credit backdrop leaves little room for error. Fitch reported that subprime auto 60+ day delinquencies stood at 6.74% at the end of 2025, compared with a pandemic low of 2.58% in May 2021. As lenders lean on automation to move faster in a tight market, how those models are governed becomes a more urgent question, especially as legislative changes may be on the horizon resulting from the mid-term elections.

Existing Rules Already Apply

Many lenders treat AI as an unmapped regulatory frontier, but existing laws cover many of its current uses in financial services. For example, the Equal Credit Opportunity Act (ECOA) and Regulation B (12 CFR 1002.9(b)(2)) require creditors to provide specific reasons for the action taken, regardless of whether the technology used was rule-based, AI, or other technology.

Interagency guidance issued in June 2023 adds that institutions should manage risks from third-party relationships at every stage, from planning through termination. That matters in auto finance, where many lenders depend on vendor models they did not build.  Though new guidance was proposed on September 15, 2026, the 2023 guidance remains very much in effect.

The complexity of auto lending raises additional challenges as loans move through dealer channels, decisions combine bureau data, alternative data and vendor scores, and customers expect fast answers. When a model incorporates dozens of inputs, producing a clear reason for a decline requires planning careful and thoughtful product design.

States Are Moving Toward Targeted Rules

State activity has shifted from broad governance mandates toward narrower, point-of-use requirements. Colorado repealed and replaced its 2024 AI law in May 2026 with a framework built around automated decision-making technology. Beginning January 1, 2027, it generally requires notice before covered tools influence consequential decisions, and a plain language explanation within 30 days of an adverse outcome.

Texas took a different route. Its responsible AI law, effective January 1, 2026, prohibits deploying AI with intent to unlawfully discriminate, states that disparate impact alone is not sufficient to show unlawful behavior, and offers a 36-month regulatory sandbox for testing new models.

The result is an evolving patchwork. A lender operating across several states may soon face notice requirements in one and intent based standards in another, which makes a single, well documented internal standard more practical than state by state fixes.

The Model Governance Challenge

In April 2026, the Federal Reserve issued updated model risk guidance, replacing guidance that dated to 2011. The update favors principles over prescriptive language and encourages treating development, validation, deployment and monitoring as a continuing discipline rather than a one-time exercise.

It also places generative and agentic AI outside its scope, even as those tools move into chat, document review and collections workflows. Institutions are expected to apply existing risk practices to them, with little direction on how. Lenders that are not directly covered by the guidance can still use it as a benchmark. The practical effect is that the newest tools carry the least specific supervisory guidance, so lenders must set their own standards.

Fraud adds pressure because criminals use the same technology to build synthetic identities and falsify income documents, pushing lenders toward AI based detection. Those detection tools are models too and need validation, because false positives can wrongly block legitimate borrowers.

Steps That Hold Up Under Any Rulebook

Details differ by jurisdiction, but the themes repeat: explainability, notice, vendor accountability and ongoing monitoring. Several practices address all four.

  • Maintain an inventory of every model and AI tool in use, including vendor and embedded tools, with a named owner and a defined purpose.
  • During vendor due diligence, look for proof of robust model governance and of model testing Monitor models continuously for drift, performance across customer segments and manual override rates, and keep logs that can be produced on request.
  • Form a cross functional review group spanning compliance, credit, technology and operations so new tools are approved against one standard.
  • Report model performance, exceptions and complaints to senior leadership on a set schedule so oversight does not depend on a single team.
  • Set an internal standard for generative and agentic tools, covering approved uses, data limits and when a person must review the output.

Controls should also match the risks that matter most. Compliance programs built around the data breach, the dominant concern of the last decade, may underweight newer exposures such as an automated agent that errs at scale or a vendor outage that halts originations. Directing effort toward those scenarios makes compliance spending do double duty as problem solving.

Preparing for Continued Change

Regulatory uncertainty is likely to persist even after the mid-term elections. Your investments in AI compliance will also pay off in practical terms regardless of the outcome, since documentation and testing built now shorten exam preparation, vendor reviews and partner due diligence later. Lenders that can explain each automated decision, document how tools are selected and monitored, and align controls with real risks will be better positioned however the requirements evolve, and better able to treat borrowers fairly in a stressed credit market.

Related Stories:

Tom Oscherwitz is Informed’s General Counsel. He has over 25 years of experience as a senior government regulator (CFPB, U.S. Senate) and as a fintech legal executive working at the intersection of consumer data, analytics, and regulatory policy. For more visit www.informediq.com.