Fannie Mae establishes new AI governance guidelines

Fannie Mae released new guidelines surrounding the use of artificial intelligence and machine learning, joining its fellow government-sponsored enterprise in mandating new AI policy.  

Introduced last week, Fannie Mae's rules officially take effect on Aug. 6 for sellers and servicers of loans it guarantees. Similar to updates already in place at Freddie Mac since March, the new governance policy requires the establishment of internal rules regarding employee use of AI and machine learning that will protect both businesses and borrowers, and spells out plans for proper review of procedures. 

"The pace of innovation brings heightened responsibility," Fannie Mae's update said. "As AI/ML models grow more complex and more deeply embedded in critical processes, seller/servicers must ensure these technologies are deployed safely, legally, ethically and in alignment with Fannie Mae's expectations."

The new Fannie Mae business policies include the following: 

• Regular communication and training on policies to be delivered to all loan officers and other personnel who work with machine learning and AI technologies. 
• Language demonstrating understanding of existing legal and regulatory requirements about how such tools may be used. 
• A designated internal overseer to implement and review procedures at least once a year to ensure legal compliance and adherence to industry-accepted best practices.  
• Guidelines that appropriately reflect risk management activities based on the lender's tolerance levels.  
• Incorporation of AI and machine learning frameworks deemed trustworthy. 
  

Fannie Mae's new guidelines also explicitly state lenders and servicers will be held responsible for mistakes and noncompliant AI use by subcontractors and vendors and mandate appropriate supervision of such providers. 

READ MORE: FHFA plan renews its effort to grow counterparty oversight

Lenders and servicers will be expected to be able to quickly disclose the types of tools in use, their providers and the safeguards put in place to mitigate risks inherent in artificial intelligence upon inquiry from the GSE.  

How Fannie and Freddie updates differ

The update from the Washington-based enterprise, which operates under the oversight of the Federal Housing Finance Agency, largely mirrors the rules implemented by Freddie Mac last month. 

Mortgage consultants on social media assured companies already attempting to address Freddie Mac's new AI mandates will find themselves prepared to effectively manage Fannie Mae's changes. 

"Both require you to be able to answer, at any time: What AI are you using? How are you using it? What safeguards do you have in place?" said Z Technology Solutions CEO Suha Zehl.

Where differences lie between the two are the types of actions prioritized by each enterprise, described by Zehl as "prescriptive" in the case of Freddie Mac, versus "principles based" for its counterpart. 

Whereas Fannie Mae emphasizes communication and a demonstrated framework to implement and apply changes, "Freddie Mac emphasizes monitoring, controls and accountability (including indemnification)," wrote Melissa Langdale, CEO of Praxis Lending. 

Freddie Mac also noted the importance of having a chain of command with defined roles for AI administration and the segregation of their responsibilities. Its policy places responsibility for any damage incurred squarely on the AI user themselves, with language that specifically indemnifies the company from risks. By comparison, Fannie Mae makes no reference to indemnification from undesirable outcomes. 

New GSE guidelines arrive as scrutiny on vendors grows with the rapid rise of AI's abilities. In February, a Pennsylvania homeowner sued Mortgage One Funding, alleging the lender had used AI-generated marketing calls in violation of the Telephone Consumer Protection Act.