The rules were originally proposed a year ago by a group of federal agencies involved in their design, including the Office of the Comptroller of the Currency, Federal Housing Finance Agency, Federal Reserve, Federal Deposit Insurance Corp., Consumer Financial Protection Bureau and the National Credit Union Administration. 

The regulations will require mortgage originators and secondary market issuers to have procedures in place to ensure confidence in AVM estimates, protect against data manipulation and provide a backstop against conflicts of interest. They also mandate ongoing random sample testing and compliance with nondiscrimination laws. 

The final rule does not spell out specific requirements for how institutions are to structure their practices but allows each to determine their own procedures based on their size and risk profile. 

"The flexible approach to implementing the quality-control standards provided by the final rule will allow the implementation of the standards to evolve along with changes in AVM technology and minimize compliance costs," the announcement said.

Finalization of the proposed regulations comes after a comment period, with the agencies receiving approximately 50 responses from stakeholders.

The addition of nondiscriminatory policy surrounding AVM use — what the regulators referred to as the rule's "fifth factor " —  received support from many commenters but also detractors. 

"While existing nondiscrimination law applies to an institution's use of AVMs, the agencies proposed to include a fifth quality control factor relating to nondiscrimination to heighten awareness among lenders of the applicability of nondiscrimination laws to AVMs," the federal announcement said.

Supporters said nondiscrimination could be seen "as a dimension of model performance and a required aspect of quality control," adding that failing to address bias might result in " safety and soundness risk."

But public remarks also pointed to pushback involving such a mandate, with some comments suggesting documented instances of AVM bias were not prevalent. Others mentioned the fifth factor duplicated existing laws and other policies, while at the same time, provided no clear performance metric for users to determine if bias existed within data. 

Some opposed pointed to the cost of compliance and limited resources at their institutions.

"They argued that small entities do not have access to an AVM's data or methodology, are unable to validate the algorithms that AVM providers use, and lack the staff to assess the AVM models results," according to the announcement. 

Commenters also said the burden of nondiscrimination compliance should fall on the AVM providers, who often hold proprietary models. The regulators noted a number of individuals calling for the creation of a separate independent third-party nonprofit to test AVM systems to ensure compliance. Such an entity would both save lenders time and improve data quality, they said.   

In addition to mortgage originations, the policy applies to AVM use in the determination of values for loan modification requests and applications for home equity lines of credit. But the regulation exempts licensed appraisers utilizing AVMs in the process of their work.      

Use of automated models gained momentum as the government-sponsored enterprises began seeking alternative appraisal methods to address speed and costs. But their adoption previously drew criticism from the likes of CFPB, who raised concerns about potential algorithmic biases involved with any tools influencing credit decision making.