Layla Zulfa
In a digital undercurrent that is rapidly becoming a torrent, sophisticated agentic artificial intelligence (AI) systems, guided by their human counterparts, are aggressively probing the defenses of Multiple Listing Services (MLSs) for access to critical real estate data. This surge, often occurring in direct contravention of established bylaws, is driven by agents seeking to leverage AI for enhanced buyer searches, automated Comparative Market Analysis (CMA) generation, and in-depth market trend research. The sheer volume of these attempts has security firms like Solid Earth on high alert, reporting a staggering increase in daily "knocks" on MLS digital doors.
Eric Steggman, CEO of Solid Earth, a cybersecurity firm specializing in protecting sensitive data, revealed that his company’s software is currently intercepting as many as 1,200 unauthorized access attempts per day targeting MLS data. "That’s on a low day," Steggman stated, underscoring the escalating nature of the issue. "Six months ago, that number was less than 10. It has exploded." This exponential growth points to a growing disconnect between the rapid advancement of AI technology and the established, often rigid, regulatory frameworks governing the real estate industry.
The landscape of MLS data access is largely unprepared for this AI-driven onslaught. Few MLSs have formally established policies governing the use of AI, and even fewer explicitly permit the direct, automated access that these agentic assistants seek. Despite this regulatory vacuum, real estate agents and industry coaches are increasingly disseminating methods for circumventing these limitations, often through popular online platforms like YouTube.
William Zhang, an agent based in Austin, Texas, and the founder of the Real Estate AI Society, has become a prominent voice in this movement. His YouTube channel features a wealth of content detailing integrations and the creation of AI-powered seller assistants, signaling his commitment to embracing AI. Zhang’s early adoption involved manually interacting with his local MLS through a browser, a method he found to be time-consuming. However, this approach quickly drew the attention of his local MLS. "I started off basically just using Claude, [I’d] open a browser to use the MLS like a human would," Zhang explained. "But obviously that got caught on by my local MLS, and they sent me a warning about it."
The warning Zhang received pertained to allowing Claude, a large language model (LLM), direct access to his MLS account. This means that for the duration of the connection, Claude could operate with the full permissions granted to Zhang’s authorized login. This practice carries significant inherent risks, including the potential exposure of sensitive information such as lockbox access codes, private seller details, and proprietary agent data.
A parallel trend is evident in the promotional efforts of Brendan Bartic, a Colorado-based agent and founder of Listing Beast. In an August 20 video shared on his Facebook page, Bartic showcased a Claude-powered browser extension designed to automate the creation of CMAs by accessing his MLS account. His post highlighted the efficiency gains: "Instead of manually opening listing after listing, checking statuses, pulling comps, comparing property details, organizing the data and building the CMA from scratch, Claude helps work through the information for you. Hands-free. Done for you. While you move on to something else."
When a commenter inquired about the legality of this practice with his MLS, Bartic expressed a lack of concern, asserting, "Claude isn’t a 3rd party app connecting into your MLS. It’s a Chrome extension that works alongside your own already authorized login, the same way a browser tab does. Nothing new is being granted access, you’re just getting help on the data you’re already cleared to see."
However, Bartic’s assertion appears to be at odds with the terms of service of Cotality, the data provider for REColorado, which Bartic’s extension is allegedly violating. For a fee of $47, Bartic is offering his followers access to a tool that directly contravenes Section 4 of Cotality’s use policy. This section explicitly prohibits the use of "any robot, spider, web crawler… or other automatic or manual device, tool or process to access, monitor, retrieve, data mine, reproduce or circumvent any portion of the Services, or in any way circumvent the navigational structure or presentation of the Services or portion thereof." The policy further states, "You may not attempt to gain unauthorized access to any portion of the Services by any means."
Bartic’s method of using a browser extension to grant AI access to his MLS account mirrors the very action for which Zhang received a warning. The use of such extensions, while often pitched as productivity enhancers, also presents significant security vulnerabilities. Global cybersecurity firm Crowdstrike warns that "AI browser extensions can boost productivity by summarizing pages, automating tasks and interacting with websites on your behalf. But because they often require deep access to your browser and may send data to external AI services, they introduce unique security and privacy risks." HousingWire reached out to Cotality and REColorado for comment but received no response prior to publication.
The Accelerating Pace of AI and Industry Inertia
In a brief statement addressing the video and the nature of his tool, Bartic acknowledged the rapid evolution of AI: "AI is moving faster than the rules around it, and every MLS is working out where it fits. Agents should know and follow their own MLS rules as they change. I’d love to see MLSs create an approved path for agents to use AI tools, the same way they did for IDX websites."
The widespread appeal of Bartic’s offering is evident in the metrics; his video garnered over 55,000 views on Facebook. This widespread reach suggests a significant segment of real estate agents are keenly interested in leveraging AI for their businesses but may lack a clear understanding of secure implementation or adherence to MLS bylaws. Bartic’s observation about the speed of AI development is accurate; the technology is advancing at a pace that challenges the adaptability of many industries, including the traditionally cautious real estate sector, and particularly its governing bodies like MLSs.
Agent Aspirations: The Drive for Enhanced Efficiency
William Zhang articulated a clear vision for how AI can revolutionize his productivity. He expressed a desire for full read/write access to his MLS data, or at least greater flexibility, believing it would unlock unprecedented capabilities. "My main objectives are twofold," Zhang stated. "One is to make sure I follow my local MLS and their compliance requirements. And 2 is to better serve my clients. And in order to do that, I think Claude or LLMs or my AI assistants do a much better job at gathering data, generating reports, creating dashboards, finding homes that my clients need, understanding the market, keeping them posted, than I can at scale. Because they are just so much better at everything that I just described."
Alex Gustafson, founder of Oppy, a company developing AI agents, shares a similar sentiment, viewing the integration of bots as an inevitable progression, partly driven by a weariness of outdated user interfaces. "If adding friction to accessing your tools is any of your company’s focus instead of creating value for your stakeholders [regardless of vertical] you will be on the wrong side of history," Gustafson told HousingWire. "Good news is, we need more free thinking creative and passionate humans than ever and this will never change. Better news is quality of life on all sides increases as friction in administrative hoop jumping decreases. I will not miss learning UIs [user interfaces]. I do not use tools for my business which require me to use a UI."
The actions of agents like Zhang and Bartic highlight a growing tension between adhering to existing regulations and embracing technological advancements that promise significant professional advantages. The question arises: to what extent should real estate agents temper their progress to avoid violating bylaws? It is undeniable that technology has outpaced policy in this sector. Gustafson’s perspective, a form of technological advocacy for efficiency and reduced administrative burden, may be more prevalent within the industry than is publicly acknowledged.
The Unintended Consequences of AI Data Access
Justin Lundy, founder of Lundy.ai, a software company that collaborates with brokerages and MLSs, pointed out that the most significant risks associated with unregulated AI access to MLS data extend beyond mere inaccuracies in CMAs or incorrect property details. The core issue, he contends, is the compromised access itself. "The risk is it storing your credentials," Lundy explained. "Because that, then, becomes something that’s hackable, but also, these LLMs, they’re still kind of a black box."
Furthermore, Steggman of Solid Earth emphasized that agents attempting to circumvent rules often overlook the self-inflicted harm. The market data accessed by their bots is perpetually contributing to the training of large AI models, often without the agent’s full awareness. "Meta’s Muse is the big one now, and I’ve seen all of a sudden [on Solid Earth’s software] that they got an uptick on attempts to use. They have a version that’s essentially free for unlimited usage from Facebook. But they tell you right up front, we’re gonna use it to train our models," he noted.
Beyond training proprietary models, these bots gain access to data belonging to other agents. While a broker may own their specific listing data, they do not own the aggregated data of other brokers within the MLS. This information, once shared with an AI assistant’s parent model, is effectively disseminated. "You are essentially giving away their property to somebody else," Steggman cautioned.
Even with security measures in place, agents circumventing rules are finding ways to grant AI access. Steggman described instances where bots log into security platforms, initiate the process of opening MLS software like Matrix, Paragon, or Flex, and then proceed to operate autonomously. "We can’t prevent that from happening, because they’ve already been authenticated in the browser," he admitted.
Matt Fowler, former CEO of Doorify, highlighted the pervasive nature of data leakage, even in seemingly innocuous actions like sending emails. "When a Realtor sends you the daily update, it’s got the list of properties in it. This one went up, this one went down, this one is new to the market. Yeah? Oh, that’s the same thing as uploading it into ChatGPT. It’s a browser that has Gemini built into it." Fowler was referring to the integrated AI features in platforms like Gmail, which can summarize emails and attachments, often displaying this information prominently.
Fowler advocates for a more structured approach: "My thought is that we have to come up with AI usage licenses that match the use, and that’s every MLS. I just came back from CMLS [Council of MLSs annual conference] where that was 40% of the content. Like, we don’t want to slow this down, but we do want to channel it into governable methods."
Navigating the Uncharted Territory: Controlled Burns and Emerging Solutions
In response to the evolving AI landscape, Bright MLS launched an AI training program in September. According to their press release, the program aims to "emphasize practical applications, including how professionals can use AI to improve productivity, strengthen client service and adapt to changing industry practices."
Meanwhile, FlexMLS, a product of FBS, introduced a Model Context Protocol (MCP) server this spring. MCP, developed by Anthropic engineers, is an open-source solution designed to facilitate the connection of AI applications to external systems. An MCP server acts as a program that exposes specific capabilities to AI applications through standardized interfaces.
Kim Prior, Vice President for Product at FBS, explained their proactive approach: "The MLS is uniquely positioned to power AI-driven real estate experiences because it brings together comprehensive, timely, trusted market data supported by established standards, governance and permissions. We developed the Flexmls MCP Server to make that data securely accessible."
However, the development and implementation of such secure systems are complex and costly. Moreover, they do not entirely resolve the fundamental question of what data an AI assistant should be permitted to access once granted entry. The ownership of aggregated MLS data is a contentious issue. While brokerages contribute their data, the MLS owns the aggregate product. Membership in an MLS signifies agreement to these terms, yet many brokerages remain hesitant about AI’s pervasive use, relying on growing stigma to discourage its adoption by other MLS members. The ongoing Compass-CMLS lawsuit, coupled with Robert Reffkin’s statements regarding the use of agent data in recruitment, underscores a pervasive industry-wide debate about data ownership, with each entity asserting its claim.
It is easy to cast MLSs as the primary antagonists in this AI data access dilemma, particularly as some agents appear to disregard bylaws in pursuit of potential commission gains. Conversely, many agents are simply unaware of the risks, driven by a genuine desire for technologically advanced methods to enhance efficiency.
"You have to remember," Steggman reiterated, "These MLSs are not all non-profits, many of them are for-profit companies. They’re there to help the members and the people that pay them money, but at the end of the day, it’s still a business that they have to run. I think each MLS is [going to] have to set their rules of what they’re willing to allow or what they’re not willing to allow, and that is certainly, uh, an active conversation in MLS boards right now."
In the context of wildfire management, "spotting" refers to windborne embers that ignite new fires, diverting resources from the main blaze and demanding immediate, critical decisions. Similarly, the current situation with AI and MLS data presents an "ember attack" scenario. Unchecked, these technological embers can spread rapidly, igniting a cascade of unforeseen consequences. The industry acknowledges that addressing this challenge will require significant resources and that the risks are escalating. Yet, from the aftermath of such disruptions, there is potential for new growth and innovation. The consensus is that a comprehensive framework for AI integration within the real estate ecosystem is not only necessary but urgently required.
