Your Home's Listing Data Is Training AI Models. Nobody Asked Permission.

Written by:

Steve Marcinuk

Date:

21 Aug 2026

Share

Share via Facebook

Share via Linkedin

Share via X

Copy to clipboard

Every time a home goes on the market in the United States, the listing data – price, photos, square footage, descriptions – enters a multiple listing service. From there, that data gets copied. According to Tim Dain, President & CEO of NorthstarMLS, the regional MLS serving Minnesota, some MLS databases get replicated hundreds or even a thousand times through various licensing feeds. Each copy lands in a different company’s hands. And increasingly, those companies are using that data to train artificial intelligence models without any governed permission system in place.

The problem most sellers never consider: AI companies can build proprietary intelligence from listing data, and neither the seller nor the agent has any say in how that intelligence gets used or sold. Dain has spent the past nine months rebuilding the data infrastructure his organization uses to distribute listing information. His argument: if the industry doesn’t govern AI access to listing data now, brokers, agents, and consumers will lose control of value they created.

Why Copies Matter More Than They Used To

In a pre-AI world, data replication was mostly a logistics issue. A company received a copy of the MLS database to power a property search portal or an appraisal tool. The copy sat in their system and served a specific purpose.

That arrangement broke down. Once a company holds a copy, Dain explains, the controller of that copy decides whether AI gets trained on it. No mechanism in the current system prevents it. No audit trail exists. No permission layer applies.

Dain frames this as a governance gap, not a conspiracy. The companies building AI products from listing data are not doing anything technically illegal in most cases – they received the data through legitimate licensing channels. But the system was never designed for this use case.

The result: “It’s completely ungoverned access right now,” Dain says. He recently attended an AI conference with over 400 exhibitor booths and says the volume of companies building intelligence products on top of real estate data is accelerating.

What Sellers Stand to Lose

If outside companies build proprietary AI models from MLS data, those models become the default tools the industry uses. The intelligence derived from millions of listings belongs to whoever built the model. If agents or brokers want access to that intelligence, they pay for it.

“We could end up buying back the intelligence that our data creates,” Dain says.

For a seller, this could mean that the AI tools an agent uses to price a home, market a listing, or identify likely buyers were built on data that seller’s listing helped create – but the agent now pays a third party for access. That cost gets absorbed somewhere in the transaction.

The Neutrality Problem

MLSs have historically operated as neutral cooperatives. Competing brokers share data through the MLS because the system doesn’t favor one over another. Dain argues that outside AI companies follow a different incentive structure. “Our primary motive is neutrality, and outside companies’ primary motive is profit,” he says. “Profit’s not bad. It’s just not our primary motive.”

That distinction matters for consumers. A neutral system surfaces listings based on standardized rules. A profit-driven system can surface listings based on whatever generates revenue – promoted placements, preferred partnerships, algorithmic choices that serve the platform rather than the buyer.

Dain is not calling for data to be locked away. His position is that access should be governed and licensed, not blocked. “It’s not to ban people from creating intelligence from the data, it’s to understand what intelligence they’re creating,” he says.

What Governance Looks Like

NorthstarMLS is building a five-layer infrastructure designed to sit between the database and anyone accessing the data. The governance layer – the middle tier – houses a policy engine, entitlements, metering, audit trails, and compliance enforcement. Every piece of data moves through that layer before reaching an outside party.

Dain describes converting NorthstarMLS’s 91 rules into machine-readable categories: some can be enforced automatically by the system, some require multi-step workflows, and some function as attestations where agents confirm compliance. The goal is a system where access is tracked, metered, and attributed – so the MLS knows who is extracting what, how much, and for what purpose.

NorthstarMLS has filed for six patents related to the technology and licensed exclusive resale rights nationally to a company called Recore.

What This Doesn’t Solve

Governance infrastructure doesn’t undo training that has already happened. Dain acknowledges that “once data becomes a model, like an AI model or proprietary intelligence, that control may get to a point where it’s not recoverable.” AI models built on scraped or replicated MLS data in years past already exist.

For now, whether a home is listed in Minnesota or anywhere else, the data enters a system designed for human search and human use. The AI era requires different infrastructure – and most of it hasn’t been installed yet.

About the Expert: Tim Dain is President and CEO of NorthstarMLS, the regional Multiple Listing Service of Minnesota, and leads its Nexus RE data governance infrastructure project.

This article is intended for informational purposes only and does not constitute legal, financial, or investment advice. The views and opinions expressed herein reflect those of the individuals quoted and do not represent an endorsement of any company, product, or service mentioned. Readers should conduct their own due diligence and consult qualified professionals before making any investment decisions.