PRISONBREAK: When Artificial Intelligence Becomes a Weapon of Psychological Warfare
What circulates online under the name “Generative AI for Good” is not a humanitarian initiative. According to documentation gathered by researchers at the University of Toronto’s Citizen Lab, it is a component of an Israeli government-linked influence operation designed to manufacture synthetic victims of the Iranian regime, escalate pressure toward regime change, and coordinate narratives with kinetic military operations in real time. The conference held in New York in April 2026 is not a starting point — it is the visible tip of a disinformation architecture that has been operational since at least January 2025.
The Citizen Lab report, published on October 2, 2025 and designated Report No. 189, identified a coordinated network of more than 50 inauthentic X profiles conducting an AI-enabled influence operation, spreading narratives inciting Iranian audiences to revolt against the Islamic Republic. The Citizen Lab The operation was codenamed PRISONBREAK. Its authors — Alberto Fittarelli, Maia Scott, Ron Deibert, Marcus Michaelsen, and Darren Linvill of Clemson University — assessed that the hypothesis most consistent with the available evidence is that an unidentified agency of the Israeli government, or a subcontractor working under its close supervision, is directly conducting the operation. The Citizen Lab
The Synchronization Between Real Bomb and Fabricated Deepfake
The most thoroughly documented moment of the operation occurred on June 23, 2025. At 11:52 a.m. Tehran time, just minutes after Israeli bombardment of Evin Prison began, a PRISONBREAK-linked account posted an AI-generated deepfake video purporting to show a direct hit on the prison’s entrance. International Business Times The real attack on Evin ran from 11:17 to 12:18 a.m. The video was already in circulation while the bombs were still falling.
The post accumulated more than 46,000 views and 3,500 likes. Within 90 minutes, accounts in the network shifted to explicitly urging Tehran residents to march on the prison and free those held inside. CyberScoop Citizen Lab researchers were unequivocal: the exact timing of the video’s posting points toward the conclusion that it was part of a premeditated and well-synchronized influence operation, with operators demonstrating apparent foreknowledge of Israeli military plans. Ynetnews
This synchronization is not a minor detail. It represents the operational fusion of kinetic and narrative warfare — where the deepfake is not a belated propaganda product but a pre-positioned asset, ready for deployment the instant the real bomb lands.
The Technical Arsenal: From Fake Profiles to Falsified Media
AI use within PRISONBREAK extended well beyond the Evin deepfake. The network used AI to distort the lyrics of a known Iranian protest song and accompany the manipulated audio with deepfake representations of three Iranian singers: Mehdi Yarrahi, Toomaj Salehi, and Shervin Hajipour — all three with histories of imprisonment for their support of the 2022 “Woman, Life, Freedom” movement. The Citizen Lab The selection of these artists was deliberate; their recognition among the Iranian diaspora makes them vectors of credibility.
The network also published a screenshot of a supposed BBC Persian article claiming that 90 members of the Iranian regime had fled the country. BBC Persian confirmed to researchers that they had never published that article. The Citizen Lab Additionally, the network revived the symbolic protest hashtag #8PMCry, encouraging citizens to shout anti-government slogans from their balconies, accompanied by numerous manipulated videos allegedly showing mass participation. Ynetnews
Two days into the conflict, accounts urged Iranians to withdraw money from ATMs, claiming the regime was confiscating citizens’ savings. AI-generated videos showing long ATM queues began circulating, with visible distortions in human figures that digital forensics researchers identified as hallmarks of machine generation. International Business Times
The Industry Behind the Operation: Disinformation Contractors
The Citizen Lab report documented not only the network of inauthentic accounts but also the commercial ecosystem sustaining it. Meta and OpenAI flagged STOIC, a Tel Aviv-based firm believed to be working on behalf of the Israeli government. Citizen Lab additionally identified two other Israeli firms, Team Jorge and Archimedes Group, that sell disinformation-for-hire services to government clients. CyberScoop
Running parallel to PRISONBREAK, an investigation by Israeli outlets Haaretz and TheMarker exposed a separate Israeli-funded campaign that manufactured fake Persian-speaking online personas to amplify support for Reza Pahlavi, the son of Iran’s deposed Shah, as a regime-change figurehead. International Business Times The “Generative AI for Good” operation recently appearing on social media fits precisely within this framework: the use of AI-generated synthetic victims to build narrative legitimacy for Western audiences.
What the images circulating on X reveal — where multiple “testimonies” from supposed victims share the same production quality, identical shot angles, and uniform aesthetic — is not a casting coincidence. It is the signature of an industrial operation.
The First Conflict Where AI Was a Weapon, Not an Accident
The 2025 Israel-Iran conflict marked an inflection point in the history of information warfare. It was the first major military conflict where generative AI played a central role in shaping public perception, and it may be a preview of what is to come. EDMO Both sides deployed synthetic content, but with significant asymmetries in sophistication and coordination.
In 2023, an estimated 500,000 deepfakes circulated online. By 2025, that figure had reached 8,000,000 — a 1,500% increase in two years. Faf The 2025-2026 Iran-Israel-U.S. conflict represents the moment these technological and geopolitical trajectories converged with maximum force.
Generative AI has improved dramatically and is now complemented by AI-powered chatbots embedded into search results and platforms like X. These chatbots have become new layers in the information ecosystem. Grok, X’s chatbot, flagged videos of Israeli Prime Minister Benjamin Netanyahu as deepfakes, generating their own wave of confusion. Brookings
The structural problem this creates is not merely technical. As AI fakes become more common, trust erodes. People begin to question everything, even real videos and images that contradict their biases. What we’re witnessing is a kind of hyper-cynicism where even truth is suspected of being fiction. EDMO
The Double Standard Nobody Wants to Name
There is a structural hypocrisy in the coverage of these operations that deserves direct acknowledgment. Disinformation campaigns attributed to Russia or Iran receive extensive coverage, thorough analysis, and regulatory response. Operations documented with equal methodological rigor that point toward actors linked to Western governments or their allies receive significantly less public attention and almost no institutional consequences.
While Western threat intelligence companies and media outlets tend to present disinformation campaigns as primarily a tool of autocratic or authoritarian countries, researchers have warned that democratic governments and private industry are increasingly playing key roles in information warfare. CyberScoop
PRISONBREAK is the most thoroughly documented demonstration of this reality to date. It does not alter the analysis of what the Iranian regime itself did — it also deployed synthetic content during the conflict. But it demands that the standard of scrutiny be applied consistently, regardless of the flag under which the operation runs.
Artificial intelligence did not choose a side in this conflict. The humans who programmed it did.
Sources
- Citizen Lab, University of Toronto 2025 “We Say You Want a Revolution: PRISONBREAK – An AI-Enabled Influence Operation Aimed at Overthrowing the Iranian Regime” (Report No. 189) https://citizenlab.ca/2025/10/ai-enabled-io-aimed-at-overthrowing-iranian-regime/
- CyberScoop 2025 “Researchers say Israeli government likely behind AI-generated disinfo campaign in Iran”https://cyberscoop.com/citizen-lab-disinformation-campaign-israel-iran-evin-prison/
- Carnegie Endowment for International Peace 2025 “Iran-Israel AI War Propaganda Is a Warning to the World”https://carnegieendowment.org/research/2025/07/iran-israel-ai-war-propaganda-is-a-warning-to-the-world
- Brookings Institution 2025 “Generative AI as a weapon of war in Iran” https://www.brookings.edu/articles/generative-ai-as-a-weapon-of-war-in-iran/
- EDMO 2025 “The First AI War: How the Iran-Israel Conflict Became a Battlefield for Generative Misinformation”https://edmo.eu/publications/the-first-ai-war-how-the-iran-israel-conflict-became-a-battlefield-for-generative-misinformation/
- IBTimes UK 2026 “Israeli-Linked Project Reportedly Using AI to Generate Fake Iranian Victims”https://www.ibtimes.co.uk/israeli-ai-deepfake-campaign-iran-1793045
- Foreign Affairs Forum 2026 “Weaponizing Illusion: Israel(US)-Iran’s War Deepfake Campaign and the Global Crisis of Information Integrity” https://www.faf.ae/home/2026/3/19/x1
