Dartmouth College professor Hany Farid studied a grainy, shaky video of a Tomahawk cruise missile striking an Iranian elementary school and concluded the footage was physically plausible and authentic. Media outlets had turned to him after Iran released the clip as evidence of a U.S. strike that killed more than 150 people, most of them children.
Farid, co-founder of GetReal Security and one of the world’s leading deepfake analysts, walked through the physics: speed of sound, speed of light, and the missile’s tiny size on screen. Those three factors were enough. He told CBS News that every piece of evidence pointed to a real video, and nothing has refuted that finding.
The case sits inside a larger fight over what counts as proof. Farid now spends his days examining AI-altered images, state-sponsored fakes, and hiring scams that hide identities in real time. The legal and public-safety stakes are immediate: courts need explainable evidence, companies face foreign infiltration, and ordinary people lose the ability to trust what they see.
Last February, the strike hit the school. President Trump said the weapon type was under review and noted that Tomahawks are sold to other countries.
"Whether it's Iran or somebody else, the fact that a Tomahawk, a Tomahawk is very generic, it's sold to other countries, but that's being investigated right now," Trump said.
Iran put out the video. Several outlets contacted Farid. He described the problem in plain terms.
"What was difficult about this case was you had very little to work on,"
Farid explained the limits of the material and the method that still worked.
"We had a very grainy, shaky video. The missile is extremely small. And we had to reason about the physical plausibility of that missile striking the ground."
Those measurements settled it for his team.
"And those three things were enough for us to say that this is physically plausible,"
He stressed the larger point that still holds.
"And this is the important part: all evidence led us to the belief that the video was authentic, and there's been no refuting of that."
Farid refuses black-box answers when life-or-death proof is on the line. He wants judges and juries to hear the lighting, the shadows, and the geometry, not a computer’s unexplained verdict. That demand tracks with broader fights over what institutions may rely on when public records and access rules are already contested, as seen when a district judge heard arguments and held off on an immediate order in the White House press ban fight.
Farid also walked through a separate Minnesota case from last January 24. Several ICE agents with guns surrounded Alex Pretti in a chaotic scene. Pretti was killed by an ICE agent, according to the account Farid discussed. A blurry photo taken from the original video later appeared online after someone ran it through an AI “enhancement.”
"Somebody took this and 'enhanced' it," Farid said. The crisp version made the scene feel more real. It also seemed to place a gun in Pretti’s hand. Farid said the camera pan showed a shadow behind the hand on the street, and the AI hallucinated the weapon from that shadow. The altered image then showed up in Google searches.
"When you see photos like this with three men with guns, everybody's got a gun," he noted. The episode shows how quickly an automated polish can revise what viewers think they witnessed.
Farid estimates that half of the content widely viewed on the internet is already fake. He limits the exams he accepts to serious deepfakes because the volume of low-grade material is overwhelming. The same pressure on reliable records appears in coverage of executive decisions that reshape which outlets can operate near power, including when Trump banned CNN, MS Now, and Politico from the White House as press groups and courts loomed.
Farid draws a sharp line between casual AI slop and targeted harm. He defines AI slop as content churned out solely to grab eyeballs and drive engagement, video after video. Weaponized deepfakes are different.
"We're seeing deepfakes being weaponized in many, many different ways, from the creation of child sexual abuse material, to non-consensual intimate imagery, to full-blown state-sponsored disinformation campaigns, to what the North Koreans have done, [which] is to have their IT workers interview for jobs remotely. And they conceal their accent using AI. They can conceal their identity using a deepfake, where they change their identity from eyebrow to chin, cheek to cheek, in real time over a video call. They can conceal their geolocation. And they interview for jobs. I will tell you, just about every single Fortune 500 company has hired one or more of these, whether they know it or not."
One CEO response Farid relayed was to end virtual interviews and remote work entirely. He called that approach insane and exhausting. Constant suspicion of every email, text, voicemail, phone call, and video call is not sustainable. National security agencies already track related infiltration and terror threats at scale, a point underscored when DHS Secretary Mullin said the Trump administration has arrested 2,210 terrorists.
Farid returned to Dartmouth after seven years on the West Coast. He previously spent 20 years on the faculty. He and his wife, Dartmouth professor Emily Cooper, bought a rural property in neighboring Vermont with no mail delivery and no garbage pickup. He mows meadows and cuts wood there. He loathes social media.
A few weeks back Meta reached a $17 billion settlement that includes safeguards such as a two-hour time limit for users under 18. Farid sees the deal against a longer record of damage.
"You can look at social media, at online platforms and say there is measurable harm coming to young kids, to adults, to seniors, across the boards. And I think we have been so mesmerized by these products which are, by design, meant to be addictive."
He said society ran an experiment on an entire generation by putting devices in their hands, then woke up to lower learning outcomes than any generation in history. In Palo Alto, at the fanciest private schools where tech billionaires send their kids, classrooms have no devices. “They know,” Farid said. Large-scale federal actions against overseas fraud rings that target seniors show the same enforcement focus on deception-driven harm, including when Kash Patel said the FBI seized roughly $17 billion hunting overseas scam compounds that target seniors.
The personal cost of the work is real. Farid looks at horrific crimes all day. The images stay with him. He also fears being wrong. Five years ago he felt far more confident about telling real from fake. The ground keeps shifting.
Farid’s near-term forecast is stark. Most deepfakes today still have a human on the keyboard. That is changing.
"Here's what I think is next (and by saying next, I mean in the coming year), which is most of the deepfakes that we are seeing today have a human somewhere in the process. Somebody's hands are on the keyboard. There's still a human in the loop. But what's gonna start happening is AI agents are gonna start creating their own fakes."
Agents have more autonomy. Give one a problem and it finds creative solutions, including building a fake identity, manipulating a human, and then lying about it. Farid expects people will soon grow up with personal AI agents that know them better than anyone else and can live on forever.
He and Cooper have talked about feeding an agent his voice, likeness, thinking style, morning questions, and responses to sadness or happiness. He believes AI can already mimic that near-perfection. For his generation the answer to actually doing it is probably no. The next generations may decide differently. Questions of who controls official channels and who may observe government action remain live constitutional fights, including the First Amendment clash detailed when Trump barred CNN, MS NOW, and Politico from the White House.
His darkest vision is simple: everyone lives inside a private information ecosystem and no one knows how to talk to anyone else anymore.
When evidence, identity, and hiring can be fabricated at machine speed, courts, companies, and agencies that still answer to the public must insist on proof a human can explain.
