With the aim of identifying criminal suspects, U.S. police departments are increasingly relying on a controversial surveillance practice to demand large amounts of users’ data from tech companies.

So-called “reverse” searches allow law enforcement and federal agencies to force big tech companies, like Google, to turn over information from their vast stores of user data. These orders are not unique to Google — any company with access to user data can be compelled to turn it over — but the search giant has become one of the biggest recipients of police demanding access to its databases of users’ information.

For example, authorities can demand that a tech company turns over information about every person who was in a particular place at a certain time based on their phone’s location, or who searched for a specific keyword or query. Thanks to a recently disclosed court order, authorities have shown they are able to scoop up identifiable information on everyone who watched certain YouTube videos.

Reverse searches effectively cast a digital dragnet over a tech company’s store of user data to catch the information that police are looking for.

Civil liberties advocates have argued that these kinds of court-approved orders are overbroad and unconstitutional, as they can also compel companies to turn over information on entirely innocent people with no connection to the alleged crime. Critics fear that these court orders can allow police to prosecute people based on where they go or whatever they search the internet for.

So far, not even the courts can agree on whether these orders are constitutional, setting up a likely legal challenge before the U.S. Supreme Court.

In the meantime, federal investigators are already pushing this controversial legal practice further. In one recent case, prosecutors demanded that Google turn over information on everyone who accessed certain YouTube videos in an effort to track down a suspected money launderer.

A recently unsealed search application filed in a Kentucky federal court last year revealed that prosecutors wanted Google to “provide records and information associated with Google accounts or IP addresses accessing YouTube videos for a one week period, between January 1, 2023, and January 8, 2023.”

The search application said that as part of an undercover transaction, the suspected money launderer shared a YouTube link with investigators, and investigators sent back two more YouTube links. The three videos — which TechCrunch has seen and have nothing to do with money laundering — collectively racked up about 27,000 views at the time of the search application. Still, prosecutors sought an order compelling Google to share information about every person who watched those three YouTube videos during that week, likely in a bid to narrow down the list of individuals to their top suspect, who prosecutors presumed had visited some or all of the three videos.

This particular court order was easier for law enforcement to obtain than a traditional search warrant because it sought access to connection logs about who accessed the videos, rather than the higher-standard search warrant that courts can use to demand that tech companies turn over the contents of someone’s private messages.

The Kentucky federal court approved the search order under seal, blocking its public release for a year. Google was barred from disclosing the demand until last month when the court’s order expired. Forbes first reported on the existence of the court order.

It’s not known if Google complied with the order, and a Google spokesperson declined to say either way when asked by TechCrunch.

Riana Pfefferkorn, a research scholar at the Stanford Internet Observatory, said this was a “perfect example” why civil liberties advocates have long criticized this type of court order for its ability to grant police access to people’s intrusive information.

“The government is essentially dragooning YouTube into serving as a honeypot for the feds to ensnare a criminal suspect by triangulating on who’d viewed the videos in question during a specific time period,” said Pfefferkorn, speaking about the recent order targeting YouTube users. “But by asking for information on everyone who’d viewed any of the three videos, the investigation also sweeps in potentially dozens or hundreds of other people who are under no suspicion of wrongdoing, just like with reverse search warrants for geolocation.”

Demanding the digital haystack

Reverse search court orders and warrants are a problem largely of Google’s own making, in part thanks to the gargantuan amounts of user data that the tech giant has long collected on its users, like browsing histories, web searches and even to granular location data. Realizing that tech giants hold huge amounts of users’ location data and search queries, law enforcement began succeeding in convincing courts into granting broader access to tech companies’ databases than just individual users.

A court-authorized search order allows police to demand information from a tech or phone company about a person who investigators believe is involved in a crime that took place or is about to happen. But instead of trying to find their suspect by looking for a needle in a digital haystack, police are increasingly demanding large chunks of the haystack — even if that includes personal information on innocent people — to sift for clues.

Using this same technique as demanding identifying information of anyone who viewed YouTube videos, law enforcement can also demand that Google turn over data that identifies every person who was at a certain place and time, or every user who searched the internet for a specific query.

Geofence warrants, as they are more commonly known, allow police to draw a shape on a map around a crime scene or place of interest and demand huge swaths of location data from Google’s databases on anyone whose phone was in that area at a point in time.

Police can also use so-called “keyword search” warrants that can identify every user who searched a keyword or search term within a timeframe, typically to find clues about criminal suspects researching their would-be crimes ahead of time.

Both of these are because Google stores the granular location data and search queries of billions of people around the world.

Law enforcement might defend the surveillance gathering technique for its uncanny ability to catch even the most elusive suspected criminals. But plenty of innocent people have been caught up in these investigative dragnets by mistake — in some cases as criminal suspects — simply by having phone data that appears to place them near to a scene of an alleged crime.

Though Google’s practice of collecting as much data as it can on its users makes the company a prime target and a top recipient of reverse search warrants, it’s not the only company subject to these controversial court orders. Any tech company large or small that stores banks of readable user data can be compelled to turn it over to law enforcement. Microsoft, Snap, Uber and Yahoo (which owns TechCrunch) have all received reverse orders for user data.

Some companies choose not to store user data and others scramble the data so it can’t be accessed by anyone other than the user. That prevents companies from turning over access to data that they don’t have or cannot access — especially when laws change from one day to the next, such as when the U.S. Supreme Court overturned the constitutional right to access abortion.

Google, for its part, is putting a slow end to its ability to respond to geofence warrants, specifically by moving where it stores users’ location data. Instead of centralizing enormous amounts of users’ precise location histories on its servers, Google will soon start storing location data directly on users’ devices, so that police must seek the data from the device owner directly. Still, Google has so far left the door open to receiving search orders that seek information on users’ search queries and browsing history.

But as Google and others are finding out the hard way, the only way for companies to avoid turning over customer data is by not having it to begin with.