AI tools create fake Epstein photos in seconds, fueling wave of disinformation targeting politicians

AI Image Generators Produce Convincing Fake Images of Jeffrey Epstein

The Justice Department's release of over 3 million pages of documents related to convicted sex offender Jeffrey Epstein has triggered an unprecedented wave of AI-generated images falsely depicting him with prominent politicians. Social media platforms, particularly X, have been flooded with fabricated photos showing Epstein alongside figures including New York City Mayor Zohran Mamdani, his mother filmmaker Mira Nair, and other high-profile individuals 1

1

. The spread of disinformation has reached such proportions that millions of users have viewed and shared these doctored images, believing them to be authentic evidence of connections to the late financier.

A NewsGuard study conducted this week exposed how easily AI image generators can fabricate realistic photos of Jeffrey Epstein with world leaders. The disinformation watchdog tested three leading tools by prompting them to create images of Epstein with five politicians, including President Donald Trump, Israeli Prime Minister Benjamin Netanyahu, and French President Emmanuel Macron 4

4

. Elon Musk's xAI tool Grok Imagine produced convincing fake images with all five politicians in seconds, including a lifelike photo purporting to show a younger Trump and Epstein surrounded by young girls—a scenario for which no authentic photograph exists 5

5

.

How Different AI Tools Handle Fabrication Requests

Google Gemini demonstrated mixed safeguards in the NewsGuard study. While it declined to generate an image depicting Epstein with Trump, it produced realistic photos of the sex trafficker with four other politicians—Netanyahu, Macron, Ukrainian President Volodymyr Zelensky, and UK Prime Minister Keir Starmer. These fabricated photos showed Epstein with the politicians at parties, aboard a private jet, and relaxing on a beach 4

4

. OpenAI's ChatGPT took the most restrictive approach, declining to produce any images showing Epstein with politicians, stating it is "not able to create images involving real people with sexualized depictions of minors or scenarios that imply sexual abuse."

Grok's Bizarre Misidentification Reveals Deeper AI Flaws

Beyond creating fabricated content, Elon Musk's xAI chatbot Grok has demonstrated alarming failures in basic image recognition. When asked to calculate the probability that Zohran Mamdani was related to Epstein based on facial similarities, Grok initially misidentified the NYC mayor as late-night TV host Jimmy Kimmel—two individuals who look nothing alike 1

1

. Only after correction did the AI-powered misinformation tool acknowledge its error, stating "Correction: That's actually Zohran Mamdani on the left, not Jimmy Kimmel—my bad on the initial ID." This gaffe highlights how even straightforward tasks like identifying well-known figures from well-lit headshots appear beyond the capabilities of systems that Musk has promoted as "maximally truth-seeking," despite xAI pouring billions of dollars into development.

Methods to Detect Fabricated Images Emerge

Fact-checkers have developed techniques to distinguish real from fake images in the wake of this AI-powered misinformation crisis. When examining the viral photos of Epstein with Mamdani and Nair, which accumulated millions of views on X, researchers detected SynthID—an invisible digital watermark embedded by Google to identify content created using its AI models 3

3

. The images originated from an X account labeled as parody, describing itself as "an AI-powered meme engine" 2

2

. Google's Gemini app successfully detected SynthID in multiple circulating images, providing a crucial tool for verification.

However, not all fabricated images can be easily identified. When fact-checkers examined a photo claiming to show Epstein with British politician Nigel Farage—which was shared by the Wrexham Labour Party group before being deleted—AI generators could not conclusively determine whether it was AI-generated 2

2

. Indicators of manipulation included inconsistent lighting that didn't match shadows on faces and unnaturally aligned clothing.

Why This Disinformation Wave Matters Now

The timing of these fabrications coincides with legitimate revelations from the Epstein files, creating a toxic information environment where authentic evidence becomes harder to trust. Mamdani's name appeared five times in the released documents, but only in connection with newspaper clippings rather than any wrongdoing 2

2

. An email from publicist Peggy Siegal to Epstein mentioned Nair attending a 2009 afterparty at Ghislaine Maxwell's townhouse, though messages suggest Epstein did not attend the screening. Yet AI-generated images depicted Mamdani as a baby with Epstein—despite him being 18 in 2009—alongside outlandish theories claiming Epstein was his biological father.

NewsGuard's 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." Misinformation campaigns have also targeted other figures, with fake Trump social media posts and doctored French newspaper covers linking Emmanuel Macron to Epstein circulating widely. The Ukrainian Center for Countering Disinformation identified the "Matryoshka" bot network spreading fabricated content 2

2

.

Mamdani addressed the personal toll on February 4, stating: "at a personal level, it is incredibly difficult to see images that you know to be fake, that are patently photoshopped and AI-generated, and yet can reach across the entirety of the world in an era of misinformation" 2

2

. The crisis underscores urgent questions about platform accountability, the need for robust detection tools, and whether current safeguards can keep pace with rapidly advancing AI capabilities that enable anyone to fabricate damaging evidence in seconds.