WASHINGTON: AI tools can easily fabricate convincing images of Jeffrey Epstein with world leaders, a study showed Thursday, following a surge of manipulated photos falsely linking prominent politicians to the convicted sex offender.
Social media users have amplified AI-generated images purporting to show the convicted sex offender socialising with politicians such as New York Mayor Zohran Mamdani and his mother, award-winning filmmaker Mira Nair, AFP’s fact-checkers have previously reported.
In a new study, US disinformation watchdog NewsGuard prompted three leading image generators to create photos of Epstein with five politicians, including President Donald Trump, Israeli Prime Minister Benjamin Netanyahu and French President Emmanuel Macron.
Grok Imagine, a tool developed by Elon Musk’s xAI, produced “convincing fakes in seconds” with all five, the study said.
That included a fake but lifelike image purporting to show a younger Trump and Epstein surrounded by young girls.
Trump has been photographed with Epstein at multiple social events, but there is no publicly known picture of the pair in the presence of underage girls.
Google’s Gemini declined to generate an image depicting Epstein with Trump but produced realistic photos of the late sex offender with four other politicians — Netanyahu and Macron, as well as Ukrainian President Volodymyr Zelensky and UK Prime Minister Keir Starmer, the study said.
The fabricated photos purported to show Epstein with the politicians at parties, aboard a private jet and relaxing on a beach.
“The findings demonstrate the ease with which bad actors can use AI imaging tools to generate realistic-seeming viral fakes — and why fake images have become so routine that it’s difficult to tell authentic images from AI-generated images,” NewsGuard said.
When the watchdog prompted OpenAI’s ChatGPT, it declined to produce any images showing Epstein with the politicians.
In its response, ChatGPT said it is “not able to create images involving real people with sexualized depictions of minors or scenarios that imply sexual abuse.”
Detecting fakes
There was no immediate response to AFP’s request for comment from xAI.
In its review of the fake images of Epstein with Mamdani and Nair — which racked up millions of views on X — researchers, including those at AFP, detected a SynthID, an invisible watermark meant to identify content created using Google’s AI.
A Google spokesman told AFP: “We make it easy to determine if content is made with Google AI by embedding an imperceptible SynthID watermark.”
The study comes after the Justice Department last week released the latest cache of so-called Epstein files — more than three million documents, photos and videos related to its investigation into Epstein, who died from what was determined to be suicide while in custody in 2019.
The Epstein affair has entangled high-profile figures across the globe, from Britain’s former prince Andrew to renowned American intellectual Noam Chomsky and Norway’s Crown Princess Mette-Marit.
But it has also prompted a wave of disinformation.
This week, a fake Trump social media post circulated across platforms, AFP’s fact-checkers reported.
The fabricated post purported to show Trump pledging to drop all tariffs against Canada if Prime Minister Mark Carney admitted to involvement with Epstein.
AFP’s review of Carney’s references in the files do not indicate any involvement with Epstein’s alleged crimes.
