NSW's AI crackdown stumbles over genuine mural in rental listing mix-up

The New South Wales government launched a high-profile campaign against deceptive rental listings this week, announcing new penalties and protections meant to crack down on AI-manipulated property photos. In its Friday media release, officials pointed to a specific example: a Leichhardt apartment they claimed showed the telltale signs of digital fakery. The problem was immediate and embarrassing. The listing agent, IB Property, flatly denied the accusation. The photograph, they explained, showed something real: a $5,000 mural of Bondi Beach that the landlord had commissioned and paid for, painted on an internal courtyard wall visible through the unit's windows. The image had not been altered, generated by AI, or touched up in Photoshop. It was a genuine photograph of genuine artwork.

When a Guardian Australia reporter visited the property at 2/14 Norton Street, the mural was not visible from the street—which may explain why the government mistook it for a digital fabrication. But inside the unit, the artwork remained exactly where it had been placed. IB Property confirmed the mural was still there and had no plans to remove it. The property had been leased in February 2026 and had not been advertised since. The government's example of deception turned out to be documentation of a real creative investment by a landlord.

Better Regulation and Fair Trading Minister Anoulack Chanthivong did not directly address the mix-up when contacted by the Guardian. Instead, he issued a statement reframing the broader intent of the legislation. "Every renter knows the frustration of seeing a property online, only to turn up and find it looks nothing like the listing," he said. "These laws are about putting an end to misleading practices and making sure renters get a fair go." The new regulations, which take effect in early 2027, will impose fines of up to $22,000 on real estate businesses that fail to disclose digitally altered or AI-generated property images. The crackdown targets a phenomenon the industry calls "housefishing"—the use of digital manipulation to distort the true size and condition of a home.

The problem the government is trying to address is real, even if the example was wrong. Generative AI has embedded itself into property marketing across Australia, and the technology complicates what was already a murky landscape of digital enhancement. Tim McKibbin, chief executive of the Real Estate Institute of NSW, acknowledged that image manipulation in real estate predates AI. "People have used Photoshop," he said. "People were removing trees and removing telegraph poles—clearly that is false and misleading." But McKibbin draws a distinction between virtual staging—placing furniture in photos to show potential—and outright falsification. Stewart Bunn of First National Real Estate pointed to concrete examples of the latter: an LJ Hooker branch in 2024 that used AI to generate a listing featuring schools that did not exist; images showing renovations that never happened; furnishings that made rooms appear larger than they were; landscaping that existed only in code.

The human cost falls on tenants navigating tight rental markets. Roxanne Gardiner, 26, described her search for a rental in Brisbane's eastern suburbs as a gauntlet of AI-generated deception. "I was constantly getting scammed by listing photos," she said. She and her household did not realize one of the first properties they visited was AI-altered until they arrived in person. The artificial furniture bore no relation to the actual rooms and consistently made spaces look bigger than they were. Beyond staged furniture, she encountered artificial greenery and soft lighting designed to transform the character of a home. "It feels cheap, lazy and insulting," she said. She eventually stopped viewing AI-altered listings altogether, instead manually filtering them out. Within three days of looking only at unedited properties, she had signed a lease. Matt Stitt, 25, had a similar experience in south Brisbane, encountering listings where floorboards had been digitally replaced to look new and area rugs had been edited in to near-cover entire rooms. "It was such a waste of time seeing these haunted houses with a digital paint job," he said.

Research underscores the scale of the trust problem. An RMIT University report led by Dr. Leah Li found a "significant lack of sector-specific AI frameworks to guide responsible AI practice within the property profession." A 2026 Cotality study showed only 40 percent of Australians trust AI-generated outputs, compared with 62 percent in Canada, 52 percent in the US, and 46 percent in the UK. Dr. Shuai Wang of the University of Queensland explained that generative AI systems prioritize visual appeal over accuracy, which in real estate manifests as what researchers call "hallucinations"—AI-generated images introducing features that were never part of the property, fundamentally altering a prospective tenant's perception of size, layout, condition, or amenity.

The legislation also addresses a separate but related problem: excessive data harvesting from renters. The new laws will protect more than 2.3 million renters by requiring landlords and property managers to use standardized application forms and banning requests for personal photos, social media handles, tattoo details, or children's information. Sensitive identity documents such as driver's licenses and passports will only be collected from the single preferred applicant before lease signing, eliminating roughly 187,000 document collections annually. Penalties for privacy breaches will reach up to $11,000 for individuals and $49,500 for corporations. The legislation also mandates upfront disclosure of embedded utility networks, so tenants know whether they will be locked into a particular gas or electricity supplier. Leo Patterson Ross, chief executive of the Tenants' Union of NSW, said deceptive imagery could artificially inflate rental markets by inflating the number of people who appear interested in a property and increasing pressure on applicants to rush decisions or offer above-advertised rent. He argued that stricter measures were necessary beyond disclosure. "Any use of imagery that misrepresents the property should be disallowed," he said. Dr. Li suggested that a long-term solution would require national coordination combining transparency requirements, professional accountability, industry standards, regulatory oversight, and consumer education. "Trust does not come from simply adopting AI," she said. "Trust comes from transparency about how AI has been used and confidence that it is being used responsibly."