Last February, a tragedy struck when a Tomahawk cruise missile hit an Iranian elementary school, resulting in over 150 casualties, predominantly children. Initially, President Trump suggested uncertainty in the missile’s origin, questioning whether it was from Iran or another entity since Tomahawk missiles are sold internationally. Iran provided video as evidence, allegedly showing a U.S. missile.
Hany Farid, a renowned expert on deepfakes, was consulted by media outlets to analyze the video. As a professor at Dartmouth College and co-founder of the cyber security firm GetReal Security, Farid faced challenges due to the video’s poor quality. He assessed the physical plausibility of a Tomahawk missile striking based on factors like speed of sound, speed of light, and missile size. Farid concluded the video was authentic, though its authenticity adds to the growing distrust in AI’s replication capabilities.
“The ability to mimic reality perfectly makes even reality questionable,” Farid noted.
In another case, involving the killing of Alex Pretti by ICE agents, Farid illustrated how AI-enhanced images can distort facts. A blurry photo, manipulated by AI, falsely depicted Pretti with a gun due to shadow hallucination—demonstrating AI’s potential to alter narratives.
Farid highlights the prevalence of fake content online, referring to “AI slop”—a term for AI-generated content designed to capture attention. While Farid’s focus remains on significant deepfake cases, he notes their weaponization, from creating illicit materials to disinformation campaigns and identity concealment by North Korean IT workers.
The challenge extends to stop fake remote interviews, but complete reliance on in-person meetups proves impractical. Farid expressed exhaustion over constantly scrutinizing communications, fearing increased difficulty in distinguishing real from fake.
“The pressure builds…it’s exhausting,” Farid remarked.
Having returned to Dartmouth College after years on the West Coast, Farid finds solace in rural Vermont—a place without mail delivery or garbage pickup, yet peaceful. Away from professional pressures, he confronts social media’s negative impact, stressing its harm through addictive designs, now evident in deteriorating learning outcomes among youth.
Recent measures, such as Meta’s settlement enforcing time limits for minors, are seen by Farid as insufficient. He warns of a future where AI might autonomously generate fakes, predicting the rise of AI agents with creative autonomy and deception capabilities.
Farid speculates on the integration of AI in personal lives, pondering its ability to replicate human identities and interaction. Despite skepticism over AI adequately substituting human relations, future generations might view it differently.
These technological advances raise critical debates about AI safety and its societal impact—an ongoing concern for Farid.