How an alleged Israeli AI influence campaign attempted to ignite revolution in Iran

New study reveals how PRISONBREAK, a sophisticated AI-driven network believed to be linked to Israel, sought to stir rebellion against Iran’s regime during peak of June's 12-day war

A new study has uncovered details of an unprecedented influence operation that reportedly took place during the 12-day war between Israel and Iran in June 2025.
According to researchers from Clemson University and Canada’s Citizen Lab, a network of fake X accounts known as “PRISONBREAK” ran a coordinated, AI-powered campaign aimed at inciting the Iranian public to revolt against the mullah regime.
1 View gallery
צילום מתוך סרטון השפעה של PrisonBreak
צילום מתוך סרטון השפעה של PrisonBreak
A photo from the campaign
The researchers say the campaign, which began in January 2025, reached peak activity in late June during intense fighting between Israel and Iran. Its digital high point came on June 23, 2025, one day before the ceasefire — just as the Israeli Air Force struck Tehran’s notorious Evin Prison, a facility known for holding political prisoners. Israeli officials confirmed the strike, describing the prison as a “symbol of oppression” and suggesting the goal was to encourage political prisoners to escape and spark unrest among their supporters.
Moments after the real-world airstrike began, a PRISONBREAK account posted an AI-generated deepfake video purporting to show a direct hit on the prison’s entrance. The fake footage quickly spread — and was even mistakenly aired by several international media outlets. Immediately afterward, the network began pushing explicit calls for Tehran residents to march on Evin Prison and “free the prisoners,” while spreading reassurances that the area was “safe.”
The timing of the video release — simultaneous with Israel’s actual strike — led researchers to conclude that the operation’s handlers may have had advance knowledge of the IDF’s plans and synchronized their digital messaging with real-world military actions.
The network, which consisted of more than 50 fake accounts, did not limit its focus to the prison. Its broader goal, the study found, was to destabilize Iranian society across multiple fronts.
Tehran's Evin Prison after an Israeli airstrike
Two days into the war, the accounts urged Iranians to withdraw money from ATMs, claiming the regime was “stealing their savings.” AI-generated videos showing long ATM lines began circulating, one of which displayed distorted human figures — a telltale sign of digital manipulation. The network also revived a symbolic protest movement under the hashtag #8PMCry, encouraging citizens to shout “Death to Khamenei” from their balconies, accompanied by numerous manipulated videos allegedly depicting mass participation.
Some PRISONBREAK accounts even impersonated legitimate media outlets, including BBC Persian, posting fabricated screenshots suggesting senior Iranian officials were fleeing the country.
Technical indicators reinforced the researchers’ assessment that this was not an organic Iranian campaign. Posts were scheduled during Israeli working hours, disseminated mainly via desktop rather than mobile devices, and made extensive use of AI tools — suggesting a highly coordinated operation run from a command center.
While state-backed influence operations have become an expected part of modern warfare, the study highlights the fusion of AI and psychological operations as a new frontier. Still, the report stops short of definitively attributing the campaign to Israel, acknowledging alternative possibilities.
“The scale was relatively small for a nation known for its cyber and intelligence capabilities,” the researchers wrote, noting that the campaign could have been run by an anti-regime Iranian group exploiting the chaos of war — or even by Iranian intelligence itself, using the operation as a trap to identify domestic dissidents.
Comments
The commenter agrees to the privacy policy of Ynet News and agrees not to submit comments that violate the terms of use, including incitement, libel and expressions that exceed the accepted norms of freedom of speech.
""