Rumors of Benjamin Netanyahu’s death were greatly exaggerated.
Mr. Netanyahu, the Israeli prime minister, was alive and well when he filmed an address to the Israeli people and posted it to social media on Friday. But the video was rapidly dismissed by some online accounts, many with ties to Iran, as an artificial intelligence-generated fake meant to deceive the world about Mr. Netanyahu’s well-being.
Did his hands, for instance, have six fingers on them, a once-common tell for A.I. fakes? Many users seemed to spot a sixth digit and were undeterred by fact–checkers finding otherwise.
Mr. Netanyahu, appearing to recognize the power and ridiculousness of the falsehood, posted a polished video from a coffee shop two days later. He held up his hands and flashed his five fingers — a new kind of proof of life for the A.I. era.
The back-and-forth was the latest demonstration of one of the dangers that A.I. technology has posed to global affairs: Not only can A.I. fakery deceive millions online, but real videos can also be dismissed as A.I.-generated lies.
The phenomenon, known as the liar’s dividend, has come to the fore during the war in Iran. Thousands of images and videos have emerged from the conflict — many real and many generated by A.I., with the differences often nearly impossible to detect by people scrolling online. That has caused some real footage of the conflict to be dismissed online as A.I. fabrications.
Solutions seem to be scant. Last week, Meta’s quasi-independent oversight board noted that the problem existed during global conflicts and crises, including the latest in Iran. The board called for the social media giant to do more to identify deceptive A.I.-generated content circulated during armed conflicts.
“This is not a conceptual threat,” said Alberto Fittarelli, a senior disinformation researcher at the Citizen Lab, a research unit at the University of Toronto. He added that anyone who was “knowledgeable of manipulation techniques” and “ruthless enough to deploy them” would take advantage of the liar’s dividend to sow distrust around the realities of war.
“Verifying everything is incredibly exhausting, and not everyone can afford doing it,” Mr. Fittarelli said.
The liar’s dividend has played an outsize role in the lead-up to the war, as protests against Iran’s theocratic government swept the country.
One video, confirmed as authentic by The New York Times, showed a protester sitting peacefully in the street as heavily armed police descended upon him. The scene spread widely online, evoking “Tank Man,” an iconic scene of resistance from Tiananmen Square in 1989. Though the footage was confirmed from multiple angles, it was nevertheless dismissed by pro-government voices online as an A.I.-made fake.
Such claims, however false, were bolstered by the growing realization that A.I.-generated images were, in fact, circulating online at the time. Citizen Lab found in October that the Israeli government or a subcontractor had used A.I. content to encourage Iranians to overthrow their government.
“Benjamin Netanyahu having to prove that he’s alive and that his image is not A.I.-generated shows that the risk cuts both ways,” Mr. Fittarelli said. Representatives for the Israeli government did not respond to requests for comment.
As the war against Iran began in February, videos and images captured the destruction in Tehran and across the country as artillery pummeled sites.
After a missile strike destroyed a girl’s school, killing at least 175 in an apparent targeting mistake by the United States, authentic video of the destruction trickled online. Some social media users, however, claimed incorrectly that the scenes of rubble, grieving parents and mass graves were false — either A.I.-made or from previous conflicts years earlier.
The prevalence of A.I. content creates an environment of doubt that makes it easier to dismiss real documentation of civilian casualties as fake, said Mahsa Alimardani, an associate director of technology at Witness, a human rights organization working on the impact of A.I. on video evidence. She noted that Tehran shut down the internet and tried to block documentation of protester deaths in January but was now invested in detailing fatalities linked to Israeli and American attacks. “The regime is engineering the information environment,” she said, adding that it “has also seeded the very doubt now being weaponized against authentic documentation.”
The Iranian government has sustained the A.I. trickery, circulating synthetic images to underscore the high cost of war. Ms. Alimardani pointed to an image of a bloody, dusty child’s backpack posted by the Iranian embassy in Austria as one example: Though the image seemed real, it was actually made by Google’s A.I. image generator, according to the company’s own detector.
Fact-checkers have relied on supporting evidence from news sources and corroborating images and videos from the scene to help determine what is real. After Mr. Netanyahu posted his coffee shop video, for example, the cafe posted its own photos of his visit — blurrier, less posed than the prime minister’s clip, but another clue of authenticity.
Still, the skepticism persisted online, propelled in part by state-affiliated media outlets such as the Tasnim news agency. Social media users claimed that the video was derived from a photo taken in 2024 (the cafe did not open until the summer of 2025). They posted similar videos of the cafe to demonstrate how easily A.I. could generate such a clip, including shots of Mr. Netanyahu in a sports jersey and of the new leader of Iran and other heads of state replicating his movements.
Grok, the A.I. chatbot created by Elon Musk’s company xAI, then mistakenly supported those assertions on X, writing that Mr. Netanyahu’s video was generated by A.I. in a post seen more than 100,000 times.
“Classic deepfake meme,” the chatbot wrote. The company did not respond to a request for comment.
Stuart A. Thompson writes for The Times about online influence, including the people, places and institutions that shape the information we all consume.
The post Netanyahu Posts ‘Proof of Life’ Video as A.I. Sows Doubts About What’s Real appeared first on New York Times.
