World’s Top Deepfake Expert No Longer Trusts His Own Eyes
The world’s foremost authority on digital image forensics, Hany Farid, has publicly acknowledged that he can no longer trust his own eyes when it comes to identifying manipulated media. In a deeply personal profile published June 14 by The New York Times, the 60-year-old UC Berkeley professor reveals that the rapid advancement of AI-generated content has reached such alarming levels of sophistication that even the father of digital image forensics is struggling to differentiate reality from fabrication.
A Crisis of Trust
Farid, who has spent more than two decades building the field of digital forensics from scratch, described his growing unease in stark terms. “I feel like I’m going blind,” he told the Times. The admission marks a watershed moment in the AI era — the person who literally wrote the book on detecting fake images is acknowledging that the technology has outpaced even his expertise.
According to cybersecurity firm Deepstrike, deepfake content has grown at a rate of roughly 900% over the past year. Farid now estimates that nearly half of what we encounter online may already be synthetic media, generated or altered using AI. The San Francisco Chronicle, which profiled Farid earlier in the week, reported that he fields requests from newsrooms and governments around the world to verify suspicious footage — a task that has become exponentially harder.
“Looking at videos like this is sort of my life,” Farid told the Chronicle. “Some mornings I’m watching videos of people getting their heads chopped off before I’ve even rubbed the sleep from my eyes.”
The Arms Race Against AI
Farid’s career began with a pivotal epiphany in 1998 while reviewing the newly updated Federal Rules of Evidence at MIT. Rule 1001 allowed digital images to be submitted as evidence in U.S. courts for the first time. Farid recognized the profound risk: digital images are inherently malleable. He began developing techniques to detect manipulation, initially using Photoshop to alter photographs of friends and then writing software to reverse-engineer the alterations.
Over two decades, Farid built the field from scratch, developing techniques including shadow calculus, vanishing point analysis, pixel anomaly detection, and metadata analysis. Some of these methods remain effective today. “There are techniques we developed 20 years ago that still work today,” he told TechCrunch in March 2025.
But the landscape has shifted dramatically. Before 2022, AI-generated videos were riddled with inconsistencies — faces morphing between frames, warped jawlines, phantom limbs appearing and vanishing. Newer AI footage can hold identities stable across hundreds of frames, replicate the texture of authentic video, and cost as little as $5 to commission using just three to five seconds of clean audio or a few public photos.
“People used to need to know things like how to use a darkroom to manipulate media, which took years of training,” Siwei Lyu, a professor at the University at Albany and director of its media forensics lab, told the Chronicle. “But now with AI, you don’t need to know anything.”
The Commercial Front: GetReal Security
In 2022, Farid co-founded GetReal Security, a startup that translates his forensic expertise into automated detection tools. The company raised $7 million in seed funding and $17.5 million in Series A funding in March 2025, with investors including Forgepoint Capital, Ballistic Ventures, Cisco Investments, Capital One Ventures, and In-Q-Tel. Its clients include John Deere, Visa, U.S. and foreign governments, defense contractors, and news outlets including Reuters and the Associated Press.
CEO Matt Moynahan described the venture as “taking Hany and trying to create a ‘Hany service’ in the cloud,” acknowledging that Farid’s expertise simply cannot scale without technological augmentation.
The Liar’s Dividend and Democratic Erosion
Farid has been a leading advocate for the Coalition for Content Provenance and Authenticity (C2PA), which is drafting standards for marking media as authentic at the moment of capture. But he warns that even technical solutions may not be enough to counter what he calls the “liar’s dividend” — the phenomenon where the mere existence of deepfakes makes it easier for officials to dismiss real, incriminating footage as AI-generated.
In January 2026, the White House posted a digitally altered photo on X of Minneapolis civil rights lawyer Nekima Levy Armstrong after her arrest during an ICE protest. The altered image darkened her skin tone and added tears. Having the White House publish synthetic content is deeply troubling, Farid said, because it muddies the historical record and makes it harder for the public to trust any visual evidence published by government officials.
“You create instability in a society not by convincing people of a lie, but by making the signals so noisy that people don’t know what to believe,” Farid told the Chronicle. “Our new reality is that we don’t have a shared reality. And that’s exceedingly dangerous for us as individuals, institutions, societies and democracies.”
The Human Toll
The psychological weight of this work is evident. Farid and his wife, neuroscientist Emily Cooper, have begun making plans to leave Silicon Valley for a farm in rural Vermont. In fall 2026, Farid is scheduled to leave UC Berkeley, where he has served as Associate Dean and Head of School for the School of Information since January 2021, to return to Dartmouth College.
Lily Xu, an assistant professor at Columbia University, described Farid as “probably the best known and most important person on the front lines of the deepfake war.” But even the most important soldier in this war acknowledges the battle may be unwinnable through detection alone.
What Comes Next
The stakes could not be higher. The Deloitte Center for Financial Services projects that AI fraud could cost the U.S. $40 billion by 2027. The European Commission estimates that 98% of deepfakes are sexually explicit material. Legislative efforts such as the Take It Down Act, signed into law in May 2025, and the Defiance Act, which passed the Senate in January 2026, represent incremental progress, but Farid argues they apply only after harm has already occurred.
As Scientific American noted in its February 2026 interview with Farid, the fundamental question is whether detection technology can ever catch up in an arms race where the attackers have the advantage of speed and scale. Farid’s answer is characteristically blunt: if we want a world where evidence still counts, we must rebuild the rules of liability and go after the choke points that make digital deception cheap and profitable.
For now, the man who taught the world how to spot a fake is preparing for a future where even he may not be able to tell the difference. And that should concern everyone.