Court documents suggest the FBI has been using controversial geofence search warrants at a scale not publicly seen before, collecting account information and location data on hundreds of devices inside the US Capitol during a deadly invasion by a right-wing mob on January 6.

While Google receives over 10,000 geofence warrants for location data in the US a year, those covering the Capitol breach appear to have been particularly productive, apparently enabling the FBI to build a large, searchable database in its hunt for the rioters.

Geofence warrants are intended to locate anyone in a given area using digital services. Google has been the target for many geofence warrants because its location technologies, which leverage GPS, Wi-Fi, and Bluetooth signals to pinpoint a phone within a few yards, are powerful and widely used.

Investigators can and do also serve warrants on phone companies. However, cell phone towers can only locate phones to within about three-quarters of a mile. While court documents suggest that the FBI collected cell tower records for “thousands of devices that were inside the Capitol” during the riot, Google’s data offers a much higher degree of accuracy.

The use of a geofence search warrant was first reported by The Washington Post, and others have previously noted specific instances of investigations that used Google geolocation data. But WIRED has found 45 federal criminal cases that cite Google geolocation data to place suspects inside the US Capitol on January 6, including at least six where the identity of the suspect appears to have been unknown to the FBI prior to the geofence warrant. One of these involved a serving Chicago police officer.

“I’m terribly concerned about the potential for misuse of that technology,“ says Ari Waldman, professor of law and computer science at Northeastern University. “Even if I think staging a coup against a democratic government is abhorrent, it doesn’t mean that constitutional privacy protections shouldn’t be in place.”

In fact, court documents refer to two geofence warrants relating to January 6, one of which a government filing seems to say was served even as the riot was raging. They were immediately sealed and are unlikely to be made public for years. However, a close reading of hundreds of court filings reveals that both the secretive geofence warrants and further Google-focused geolocation warrants delivered a wealth of information about dozens of suspects.

Geofence warrants are essentially a fishing expedition: Investigators know roughly where and when a crime was committed, and want to find out who might have been nearby at the time. As this would normally include innocent people and bystanders, Google requires law enforcement to go through a three-step process to access the information.

A geofence warrant initially seeks an anonymized list of devices tracked within a specific area at a specific time. Investigators then use that list to focus on tracks that look suspicious, and can ask Google to widen the time or geofence boundaries on only those devices. Finally, investigators can go back to Google to unmask the real name, email, phone number, and other information of just a few account holders. Courts can and have—albeit very rarely—denied geofence warrant requests that are overly broad.

But where a typical geofence fishing expedition might catch only one or two suspects, the January 6 investigation appears to have landed a netful.