Leaked court doc shows US investigators secretly used a “keyword warrant” to order Google to provide data on anyone searching a victim's name and personal info
The U.S. government is secretly ordering Google to provide data on anyone typing in certain search terms, an accidentally unsealed court document shows. Tweets: @granick , @iblametom , and @iblametom Tweets: Jennifer Granick / @granick : Warrants for keyword searches are really scary. The tactic enables police to identify people merely based on what they might have been thinking about, for whatever reason, at some point in the past. 1/n https://t.co/sKijEEVcXn Thomas Brewster / @iblametom : UPDATED: Turns out that there were broader keyword warrants in 2018 that weren't that widely discussed. They were made during the investigation into serial bombings in Austin. Thanks to @lynch_jen for pointing me to them! Links to docs at the bottom of the story. https://twitter.com/... Thomas Brewster / @iblametom : SCOOP - In 2019, the government told Google to provide data on anyone who searched for the name of a kidnapping and sexual abuse victim, her phone number, address and her mother's name. We've seen “keyword warrants” before, but this is the broadest yet. https://www.forbes.com/...
Context & Ripple Effects
The leaked filing closes a loop that has been opening since 2018, when the FBI served Google a gagged warrant covering every user at two robbery-adjacent locations — a request Google ultimately did not fill. A year ago, [[a:1160929|court documents in a Florida case showed Google handing over IP addresses for anyone searching specific keywords]], meaning the company had already complied at least once before this accidental unsealing made the tactic publicly undeniable.
What makes this disclosure land is Jennifer Granick's framing: a keyword warrant doesn't target a person or a place, it targets anyone who typed a phrase — here, a victim's name and personal details — turning ordinary curiosity into a suspect pool. Thomas Brewster's follow-up note that broader keyword warrants existed quietly since 2018 suggests the practice predates its public debut.
First-order effects
- Anyone who searched the victim's name and personal info during the warrant window became an identifiable suspect in federal hands, with Google compelled to surrender their account data under seal.
- Because the document was only accidentally unsealed, other keyword and geofence orders against Google may still be active and unknown to the public — the secrecy mechanism, not just this warrant, is exposed.
Second-order effects
- The tactic scales beyond violent cases: Bloomberg's review of 115 warrants across five states found investigators leaning on Google location and search data increasingly for nonviolent investigations, normalizing dragnet queries as routine evidence-gathering.
- Every unsealed example raises pressure on Google's compliance posture, which already spans spectrum — refusing the 2018 geofence while fulfilling the Florida keyword request — and forces the company to defend where it draws the line publicly rather than behind gag orders.
Third-order effects
- If the trajectory holds — from geofenced places to searched words to, per later coverage, a state supreme court letting police take Google search data without any warrant at all (Pennsylvania's ruling) — the legal default shifts toward treating query logs as presumptively available to police.
- Structurally, the burden of privacy protection migrates from warrants to platforms and users: whoever controls the search index holds the power to identify thought patterns at scale, making platform policy and encryption choices de facto civil-liberties policy.
The trend: Law enforcement is moving up the data stack — from physical locations to search terms themselves — turning what people type into a standing investigative instrument whose legal limits courts are still drawing.