Google sued after ai search doxxes epstein survivors from botched us file dump

The Justice Department dumped 2.3 million pages of Jeffrey Epstein documents online in December, promising transparency. Within hours, Google’s ai-powered search engine was serving up the full names, phone numbers, hometowns and even family photos of women who say they were raped as minors. Now one of those women, listed only as Jane Doe 1, is suing both the federal government and Alphabet in a class-action complaint that accuses Washington of negligence and Mountain View of profiting from it.

The suit, filed late Monday in San Jose, is the first to test whether the 1974 Privacy Act applies to algorithmic search results. It also lands at the worst possible moment for Google, which is already fighting California’s new ai safety bill and a pending EU rule that would force real-time takedowns of non-consensual intimate imagery.

Redaction never happened

According to the 42-page complaint, DOJ contractors ran a keyword search across the Epstein archive, slapped a ‘‘public’’ tag on every hit and uploaded the lot to the transparency portal without opening a single PDF to check for Social Security numbers, passport images or nude photographs. The result: spreadsheets labelled ‘‘victim contact list’’ and interview transcripts in which minors’ names appear in the first line. DOJ quietly pulled some batches two weeks later, but cached copies – complete with original file paths – remain reachable through Google’s ‘‘ai Overview’’ snippets.

‘‘They adopted a publish-now, redact-later workflow that made doxxing inevitable,’’ the complaint reads, quoting internal Slack messages between contractors who joked about ‘‘Epstein Easter eggs’’ in the data dump.

Google’s refusal to forget

Lawyers for Jane Doe 1 sent Google 27 removal requests under California’s ‘‘Eraser Button’’ law between January and March. The company’s trust-and-safety team denied each one, arguing that the files are ‘‘official court records’’ and therefore exempt from de-indexing. The lawsuit counters that once Google’s generative model synthesises the data into an answer box, it stops being a passive host and becomes a publisher – one that is ‘‘monetising trauma with every ad click’’.

In a sworn declaration, the plaintiff says she has received ‘‘dozens of sexually explicit messages’’ from strangers who found her Instagram handle through an ai Overview that auto-completes the search ‘‘Epstein victim [her first name] contact’’.

What the plaintiffs want

The proposed class covers ‘‘all persons identified by name in the Epstein files whose personal data appears in Google ai results’’ – potentially thousands of women. They seek a court order forcing Google to purge every cached URL, statutory damages of $7,500 per violation under the California Consumer Privacy Act, and a punitive award tied to Alphabet’s ad revenue from searches containing the word ‘‘Epstein’’ since December. The federal defendants face separate claims under the Privacy Act, where statutory damages cap at $1,000 per person but can be multiplied if the court finds ‘‘wilful disregard’’.

Google declined to comment on pending litigation. A DOJ spokesperson said the department is ‘‘reviewing redaction protocols’’ but gave no timeline for re-issuing a sanitized archive.

The case is Jane Doe 1 v. United States et al., 5:24-cv-01892, Northern District of California. A hearing on the motion for class certification is set for October. Until then, a Google search for ‘‘Epstein victim list’’ still returns a neat, AI-generated carousel of names and phone numbers – and the clock on the statute of limitations is ticking for every survivor who has not yet come forward.