Real estate AI: high stakes, high regulatory exposure
Access to housing is one of the most fundamental decisions in people's lives, and AI is increasingly embedded in every stage of the housing market: property discovery and recommendation, rental application screening, mortgage underwriting and credit assessment, property valuation, and investment portfolio management. Each of these AI applications creates specific regulatory exposure, much of it under laws that predate AI but apply with full force to algorithmic systems.
Automated valuation models and EU AI Act
AVMs used in mortgage underwriting — models that generate property values to support lending decisions — fall squarely within EU AI Act Annex III Category 5: AI systems intended to be used for evaluation of the creditworthiness of natural persons or to establish their credit score. This categorisation means full high-risk AI compliance requirements apply: conformity assessment, technical documentation, data governance, human oversight, and post-market monitoring. The December 2027 deadline under the Omnibus applies.
For real estate lenders using AVMs, the EU AI Act compliance timeline needs to be integrated with existing model risk management practices. Most sophisticated lenders already conduct AVM validation — the EU AI Act builds on this with additional documentation, oversight, and transparency requirements that go beyond typical internal MRM programmes.
Tenant screening: fair housing exposure
AI tenant screening tools — systems that assess rental applications and recommend approval or rejection — have attracted significant regulatory scrutiny in the United States. HUD's fair housing enforcement position is unambiguous: the Fair Housing Act applies to AI screening decisions. Landlords who use AI screening tools that systematically reject applicants from protected groups face Fair Housing Act liability regardless of whether the algorithm explicitly uses protected characteristics. Proxy discrimination — systematic rejection of applicants based on factors that correlate with race, national origin, disability, or other protected characteristics — is unlawful under the Act.
In practice, this means real estate operators using AI tenant screening must: independently assess screening tools for disparate impact across protected groups before deployment; monitor screening outcomes for discriminatory patterns on an ongoing basis; ensure individual applicants can obtain meaningful explanation of adverse screening decisions; and maintain human oversight for final tenancy decisions. "The algorithm rejected you" is not a legally sufficient explanation for a rejected housing application.
Algorithmic rent pricing and antitrust risk
Perhaps the most significant emerging legal risk in real estate AI is algorithmic rent pricing — the use of AI platforms (such as RealPage) that provide landlords with pricing recommendations based on aggregated market data. DOJ and FTC investigations have focused on whether these platforms facilitate coordination among landlords who would otherwise compete on price — a potential violation of antitrust law even without any explicit agreement between competitors.
The theory: if competing landlords all accept pricing recommendations from the same AI platform trained on data from all those landlords, the AI may be effectively coordinating pricing in ways that harm tenants — without any communication between the landlords. Whether this constitutes unlawful price coordination under antitrust law is being litigated, but the risk is real and the regulatory attention is serious. Real estate investment companies using third-party pricing AI should monitor this litigation and consider whether their use of such tools creates antitrust exposure.
