Hany Farid, a leading expert in digital forensics and artificial intelligence (AI) deepfake detection, finds himself increasingly grappling with the challenge of verifying the authenticity of online videos amid the rapid advancement of AI-generated content. Based in Berkeley, California, Farid has spent over 20 years developing tools and methods to discern real images and videos from fabrications. However, he now confronts a growing reality where deepfakes have become so sophisticated that even he struggles to confidently separate fact from fiction.

The issue came into sharp focus when Farid was asked to analyze a viral video depicting what appeared to be a U.S.-made missile striking an elementary school in Minab, Iran, resulting in the deaths of over 150 people, many of them children. The clip, shared widely on social media and published by an official Iranian news agency, raised urgent questions about its authenticity. Using a range of forensic techniques—including geolocation, shadow analysis, pixel measurement, and frame-by-frame tracking of the missile’s trajectory—Farid and colleagues found no conclusive evidence that the video had been manipulated. Still, despite their verification efforts, Farid hesitated to declare it definitively real, highlighting the pervasive uncertainty that now defines his work.

This incident exemplifies the broader challenge posed by the proliferation of AI-generated images, audio, and videos that can convincingly mimic reality. Farid reports receiving multiple daily requests from governments, journalists, law enforcement agencies, and human rights groups seeking help in verifying digital content amid what many describe as an “infodemic” of misinformation. Alongside this professional strain, Farid has personally experienced the risks of deepfake technology, having his own voice replicated to impersonate him in fraudulent communications.

Farid and his wife, Emily Cooper, a vision scientist at the University of California, Berkeley, have decided to relocate from the Silicon Valley area to rural Vermont as they seek respite from the digital inundation and its psychological toll. The couple has purchased a cabin near Dartmouth College where they plan to continue their research remotely and find balance through a closer connection to nature.

In his lectures to students, Farid underscores the accelerating speed at which deepfakes are evolving and the consequential threats they pose to democracies, public trust, and social stability. He warns that as AI’s ability to seamlessly fabricate visuals and audio improves, traditional methods of verification become less effective, and the delay in detecting falsified content means misinformation can solidify before truth emerges.

Farid’s tools, which include algorithms measuring physiological signs such as pupil dilation and blood flow to detect synthetic avatars, remain effective but time-consuming. This timing discrepancy exacerbates the problem, as misinformation often spreads faster than investigations can be completed.

Despite the growing challenges, Farid maintains cautious optimism about the potential for solutions, emphasizing the need for public awareness and technological innovation. At the same time, he acknowledges the uncertainty and complexity of the path ahead. As the digital landscape shifts, his work and personal life reflect the urgent need to adapt to a new reality in which seeing is no longer believing.