Cops can now find your home from a photo of your washing machine

The demo photo is a washing machine and a dryer. No GPS tag, no metadata, no street sign through a window. Just two appliances in a laundry room, the kind of photo nobody would think twice about posting.

Upload it to a tool called Raven, pick "Find Property," choose a region, and it comes back with a candidate: an Airbnb listing in Charlotte, North Carolina, whose photos show the same room.

That is the pitch Graylark Technologies is making to police, according to a report published October 12 by Joseph Cox at 404 Media. Raven, the company says, can take a single interior photo and identify the specific house, apartment or short-term rental where it was taken, using only what is visible in the frame. 404 Media reports the tool is "clearly pulling data from sites like Zillow and Airbnb."

From GeoSpy to Raven

If the name Graylark rings a bell, it is because the tool used to be called GeoSpy. In its earlier life it geolocated outdoor photos: upload a street scene and it would guess where in the world it was taken, often scarily well. It was open to the public. Then 404 Media found people in the product's Discord asking for help finding and stalking women, and Graylark locked access down to government and enterprise customers.

404 Media has previously reported that the Los Angeles Police Department and the Miami-Dade Sheriff's Office bought access. Gadget Review, citing 404 Media, says Miami-Dade's Cyber Crimes Bureau has been piloting it, using a model trained for the county. Graylark raised $10.7 million in seed funding led by General Catalyst, according to Gadget Review.

The rebrand to Raven comes with a new trick. Outdoors, there are mountains, road markings and architecture to go on. Indoors there should be much less. That is what makes the interior feature unnerving. Kitchens and bathrooms all look alike, until a model has seen enough listing photos to know they don't.

How it works, as far as anyone can tell

Raven is not a reverse image search. It does not need your exact photo to exist online. According to reporting from Gadget Review and AI Weekly, both based on 404 Media's story, it picks out distinctive details such as flooring, cabinetry, appliance placement and room layout, then hunts for listing photos of the same property, even when those photos were taken from a different angle years apart.

The text in Raven's results gave away part of the source. 404 Media found that the language in some results directly quoted Airbnb and Vrbo listings. Another demo appeared to use Zillow imagery. Graylark has not laid out its data pipeline in public.

The CEO, Daniel Heinen, frames the tool around trafficking. Short-term rentals get used in trafficking and exploitation cases, he wrote, and for investigators, matching a photo to a listing "can turn an unidentified room into a location lead they can verify and pursue."

He also admitted the capability was something of an accident. According to coverage of the 404 Media piece, Heinen said a bathroom selfie of his girlfriend was matched to a location in San Francisco. "We didn't train it to do that. That was kind of an 'Oh shit' moment for us," he said.

That quote is the whole story in one line. A company built a model, fed it the internet, and discovered it could find people's homes from a mirror selfie. Then it sold the feature to the police.

The platforms say they had nothing to do with it

Airbnb told 404 Media it has no connection with Graylark, that its terms of service ban scraping, and that it is looking into the matter. "Protecting our community's privacy is important to us," the company said, adding that it strips location metadata from listing photos.

That last point is a little beside the point now. Stripping the GPS tag from a photo used to be the privacy fix. Raven doesn't need it. The location is in the tile pattern and the shape of the shower.

Zillow said scraping violates its terms and that it pursues "legal avenues" when needed. Neither company has said it is taking action against Graylark specifically. Vrbo did not respond, according to Gadget Review.

Does it even work?

Here is the inconvenient bit for the sales deck. A Miami-Dade spokesperson, Joseph R. Peguero Rivera, told 404 Media that the agency's use of GeoSpy "has been limited and largely exploratory" and "has not resulted in any significant investigative breakthroughs or arrests."

The tool also only works where there is something to match against. If your apartment has never been listed for rent or sale with photos, Raven has nothing to compare your bathroom to. That protects a lot of homes. It protects far fewer in big rental cities, where nearly every unit has been photographed for a listing at some point.

Accuracy is its own question. A tool that returns a "candidate property" will sometimes be wrong, and the cost of being wrong here is a police visit to the wrong door. Gadget Review notes Graylark pitches Raven as a lead generator and "not a substitute for corroboration." That framing has a history. It is how facial recognition was sold too, and the ACLU has counted more than a dozen wrongful arrests tied to it.

Your old listing photos are now evidence

There is no clear answer yet on what legal process, if any, police need to run a photo of someone's living room against a scraped database of rental listings. There is no warrant requirement for looking at a public Airbnb page. Combining millions of them into a machine that turns any interior photo into an address is a different thing, and the law hasn't caught up.

For everyone else, the practical takeaway is uncomfortable. Every time a landlord, a host or a real estate agent photographed the place you live, they were building a reference image for a system like this. You didn't consent to that, and you can't take it back. The washing machine was enough.

Sources: 404 Media, Gadget Review, TechSpot, ACLU.