In late January the US Department of Justice released more than 3.5 million documents tied to convicted child sex offender Jeffrey Epstein. The archive — which also contains over 180,000 images and more than 2,000 videos — lets the public examine Epstein’s connections to celebrities, business figures and other powerful people. But the release also triggered a wave of rumors, misinterpretations and deliberate disinformation, making it hard for many to separate verified fact from viral fiction.
“Just having the documents doesn’t mean you can search them immediately,” said Gianna Grün, head of data journalism at DW. Documents must be made machine‑readable before they can be effectively searched and analyzed. Within hours of the DOJ upload, social platforms filled with screenshots, name lists and sensational claims — not all of which withstand scrutiny.
Even though the materials are public records, they are not automatically verified, cautioned Steve Eder, an investigative reporter at the New York Times. A name appearing in a document does not by itself prove involvement in criminal activity: every mention must be examined in full context.
Zuckerberg image claimed to show Epstein
Claim: A widely shared post on X (formerly Twitter) purportedly showed Meta CEO Mark Zuckerberg seated beside Epstein, with a young woman leaning on Zuckerberg’s lap; LinkedIn co‑founder Reid Hoffman was also pictured.
Fact: Fake. The viral image carries a “DFF” watermark and traces to an X account that regularly posts AI‑generated images and videos. The viral picture blends authentic elements with synthetic ones. A real photo of Zuckerberg is included in the DOJ release — reportedly from an August 2015 dinner that also included Elon Musk — but Epstein is not in that image. Zuckerberg has said he had no contact with Epstein beyond that dinner. Hoffman says he met Epstein through a fundraising relationship, which he now regrets.
Analyzing millions of documents takes time
Newsrooms are working through the material; some collaborate across outlets, others use AI tools to structure the data. The New York Times’ AI project editor Dylan Freedman described building a tool that used the DOJ’s search functionality to extract pages of results into a spreadsheet. Reporters then populated tabs for key figures, linked results back to source documents and crowdsourced verification, checking each mention in its surrounding context.
Tom Hanks denied entry into Greece?
Claim: A viral post claimed actor Tom Hanks was refused entry into Greece after his name appeared in the Epstein files, citing a non‑existent “Greek Foreign Minister Jostaki Barronopolous” allegedly revoking Hanks’ citizenship.
Fact: False. The story is fabricated. Hanks received Greek citizenship in 2020; the minister named in the post does not exist (the current foreign minister is George Gerapetritis). Hanks’ name does appear in the files, but only in passing references with no indication of wrongdoing.
Context matters
DW’s data journalist Grün stressed that context is crucial: the number of times a person’s name appears does not tell the whole story. A mention could come from a mailing list, a third party’s comment, a routine newsletter or an email chain — and each instance requires semantic and situational examination to determine relevance.
A process that may take years
The DOJ has identified more than six million potentially responsive pages linked to the Epstein files; roughly half have been released so far. Given the volume and complexity, experts expect that verifying, contextualizing and investigating the documents will take years. Dozens of newsrooms are still indexing, analyzing and corroborating material, and they continue to warn the public against drawing quick conclusions from isolated screenshots or name lists.
Edited by: Sarah Steffen, Rachel Baig