On June 6, 2024 federal regulators published a final rule addressing Quality Control Standards for Automated Valuation Models. This was a collaborative rulemaking effort that included regulators from the Office of the Comptroller of the Currency, the Federal Deposit Insurance Corporation, the Federal Reserve, the National Credit Union Administration, the Consumer Financial Protection Bureau (“CFPB”) and the Federal Housing Finance Agency. The rule takes effect twelve months from the data of publication in the Federal Register, and as of the date of this Cabinet, the final rule has still not been published in the Federal Register. As the CFPB explains in a blog post entitled “CFPB Approves Rule to Ensure Accuracy and Accountability in the Use of AI and Algorithms in Home Appraisals” regarding the final rule, the new rule, which is largely consistent with the proposed rule published last year, “requires companies that use . . algorithmic appraisal tools to put safeguards into place to ensure a high level of confidence in [home] value estimates, protect against the manipulation of data, avoid conflicts of interest, and comply with applicable nondiscrimination laws.”
Stemming from Dodd-Frank amendments to the Financial Institutions Reform, Recovery, and Enforcement Act of 1989 (“FIRREA”), this new rule applies to “automated valuation models” (“AVMs”) that are used to “determine the collateral worth of a mortgage secured by a consumer’s principal dwelling.” The scope of the rule covers both mortgage originators that use AVMs for credit decisions, but also secondary market issuers (i.e., “any party that creates, structures, or organizes a mortgage-backed securities transactions) who use AVMs to make certain determinations regarding securitizations, and the rule is focused upon ensuring that the AVMs used meet certain quality requirements. The secondary market issuer activity that is covered is called a “covered securitization determination”, which is defined to mean “a determination regarding: (1) Whether to waive an appraisal requirement for a mortgage origination in connection with its potential sale or transfer to a secondary market issuer; or (2) Structuring, preparing disclosures for, or marketing initial offerings of mortgage-backed securitizations.” The Federal Register commentary generally identified the GSEs as being subject to these requirements, but the requirements would apply to any secondary market issuer. Significantly, certain AVM uses are not covered by the rule, specifically when AVMs are used solely to review completed determinations and also when licensed appraisers use AVMs to prepare an appraisal. Note, however that non-licensed individuals who prepare evaluations of home value, but not appraisals, must ensure their AVMs meet the quality requirements of the rule.
In terms of the quality requirements, the final rule adopts what the regulators describe as “a flexible approach to implementing the quality control standards [that would] allow the implementation of the standards to evolve along with AVM technology and reduce compliance costs.” This approach does not seek to foster uniformity in the AVM market. Instead, “[t]he quality control standards adopted are clear and simple” and are consistent with the “four quality control factors from the statute”, but also include a fifth factor not in the statute, but implied, the regulators argue, by other laws, which is to ensure AVMs protect against discrimination. As such, mortgage originators and secondary market issuers using AVMs for a covered reason must “adopt and maintain policies, practices, procedures, and control systems” to ensure that the AVMs adhere to the following quality control standards: (1) That there is a high level of confidence in the estimates produced; (2) Protection against the manipulation of data; (3) Avoidance of conflicts of interest; (4) Required random sample testing and reviews; and (5) Compliance with applicable nondiscrimination laws.
On the technology side of things, the regulators note that some valuation products in the marketplace that do not “currently meet the definition of an AVM may meet that definition in the future” through the use of artificial intelligence, machine learning and other technologies. Accordingly, all valuation products generally (i.e., not just those that are technically AVMs) that are (1) automated; (2) a model, as defined by the Interagency Supervisory Guidance on Model Risk Management; and (3) designed to estimate the value of a consumer’s principal dwelling collateralizing a mortgage, are subject to the quality control requirements.
Nota bene -- A quick aside on the use of the term “AI” by the CFPB (and by yours truly, at least in the title of this piece) in describing the algorithms that are used in typical AVMs – just because software uses algorithms does not necessarily mean that the software is “artificial intelligence” in the sense that is being marketed everywhere these days. The financial services industry has used algorithms for decades, for all sorts of purposes. It is important that folks (and regulators) understand that just because algorithms are involved, that does not mean that they constitute “AI”, necessarily. Of course, do notice that the CFPB phrased the blog post title so that it suggests the rule addresses algorithms and AI, and not just AI, so presumably they know the difference.