Key Takeaways
- Russian units and milbloggers are increasingly deploying AI‑generated or AI‑altered videos to falsely claim territorial gains that have not been achieved on the ground.
- The Institute for the Study of War (ISW) identified at least 35 suspect videos between June 1 and August 30 2024, with commercial AI‑detection tools rating ≥ 10 of them as highly likely manipulated.
- Common visual flaws include distorted soldier proportions, inconsistent shadows, unnatural movements, and flags that appear to disappear or deform; however, accurate geolocation alone does not guarantee authenticity.
- Notable fabricated clips purport to show the “liberation” of Bely Kolodez (Kharkiv), Mykolaipillia (Donetsk), and Sviatohirsk (Donetsk), all of which remained under Ukrainian control at the time of the claims.
- Pro‑war Russian channels such as Rybar and Voenny Osvedomitel have publicly criticized the practice, warning that false reports can distort commanders’ battlefield understanding and risk flawed operational planning.
- The AI‑driven disinformation campaign serves both domestic and international audiences, aiming to reinforce the Kremlin’s narrative of initiative and to make exaggerated claims appear credible.
- ISW expects the use of AI‑generated material to expand as video‑generation technology becomes more accessible, and warns that future tactics could include fabricated satellite imagery; meanwhile, visual anomalies remain a reliable cue for open‑source investigators.
Overview of the ISW Findings
The Institute for the Study of War (ISW) released a report detailing how Russian sources are exploiting artificial intelligence to manufacture false evidence of battlefield success. According to the think‑tank, the practice “does not yet appear to be centralized across the Russian military. Instead, several Russian units and milbloggers appear to be producing or amplifying AI‑altered footage, potentially to conceal battlefield failures or create the impression that units have met Kremlin‑set objectives.” This decentralized approach allows a variety of actors—from frontline units to influential Telegram channels—to push a unified narrative of Russian advance, even when the front lines remain static or are shifting in Ukraine’s favor.
Scale of AI‑Altered Footage and Detection Challenges
ISW’s analysis covered the period from June 1 to August 30 2024, during which it flagged at least 35 videos suspected of being AI‑generated or AI‑altered. Commercial AI‑detection software assessed with high confidence that a minimum of ten of those clips had been manipulated using AI models. The report cautions that “the actual number is likely higher, because existing detection tools don’t consistently identify AI manipulation.” Typical anomalies noted by analysts include “distorted soldiers, inconsistent shadows, unnatural movements and flags that appear to disappear or deform.” Additionally, creators often employ “camera movements, zooms and jump cuts” to obscure these tell‑tale signs, making visual verification more difficult for casual observers.
Case Study: The Bely Kolodez Flag‑Raising Video
On July 27, the milblogger DIVGEN posted a video depicting a Russian soldier waving a flag over what was claimed to be the newly liberated settlement of Bely Kolodez in the Kharkiv region. The caption read: “Own sources. Soldiers of the 3rd Motorized Rifle Battalion of the 127th Regiment, together with ‘Storm‑V’ detachments, have completed the liberation of the urban‑type settlement of Bely Kolodez [in the Kharkiv region].” ISW concluded that the footage was an AI‑generated fake, citing “the soldier’s implausible body proportions and his sudden disappearance from the bushes when the drone camera zoomed out to show the supposedly captured settlement.” These visual inconsistencies—particularly the soldier’s abrupt vanishing act—served as red flags that the scene had been synthetically constructed rather than filmed on location.
Case Study: The Mykolaipillia “Growing Flag” Clip
Another prominent example emerged on August 20 from the Telegram channel “Yuzhny Rubezh” (Southern Frontier). The video showed a Russian flag seemingly “growing” out of the roof of a house before retracting back into it. The accompanying claim asserted that “the village of Mykolaipillia in the Donetsk region had been ‘liberated as a result of active and decisive actions by a unit of the Southern Group of Forces.’” ISW’s analysts noted the flag’s unnatural morphing behavior as a hallmark of AI manipulation, yet the clip was widely circulated within pro‑Russian circles as proof of territorial control.
The Sviatohirsk Controversy
Perhaps the most high‑profile case involved the town of Sviatohirsk, a strategic point guarding the northern flank of the Sloviansk‑Kramatorsk “fortress belt” in Donetsk. In early September, Russian leader Vladimir Putin and the Russian Defense Ministry announced that Russian forces had captured the town. However, Brigadier General Andrii Biletskyi, commander of Ukraine’s 3rd Army Corps, published a video address from the town centre, demonstrating that Sviatohirsk remained under Ukrainian control. Pro‑war Z‑channels initially amplified the Russian claim, posting two videos purportedly showing the “liberation” of Sviatohirsk. The “Voenny Osvedomitel” (Military Informant) channel later admitted that the footage had been generated by a neural network.
The influential pro‑war Rybar channel also weighed in, criticizing the trend: “such ‘deliberately false’ reports could expose serious problems in Russia’s battlefield reporting.” Rybar warned that false reports about captured settlements “could distort commanders’ understanding of the battlefield, potentially leading to military operations being planned based on an inaccurate tactical picture.” The channel further noted that Ukrainian footage showing Biletskyi walking around the Sviatohirsk Lavra directly contradicted the Russian narrative of control.
AI Fakes as Tools to Exaggerate Advances
ISW observed that fabricated footage is especially valuable for Russian units that have stalled on the front lines. For instance, milblogger Voin DV repeatedly attributed territorial gains to Russia’s 36th Motorized Rifle Brigade using clips that ISW assessed or suspected were AI‑altered. Similar tactics appeared along the Donetsk “Fortress Belt,” where Russian forces have endured slow, costly progress. After Ukrainian President Volodymyr Zelensky announced in August that Ukrainian forces had liberated 26 settlements in the Oleksandrivka direction, Russian sources responded by posting footage purporting to show Russian troops still occupying several of those localities. ISW assessed that some of this counter‑claim footage had also been altered using AI, illustrating how the disinformation cycle works in both directions.
Targeting the Information War Abroad
The AI‑driven deception is not confined to domestic consumption. ISW emphasized that “the fabricated videos are intended not only for Russian audiences but also to shape perceptions of the war abroad.” Russian officials have long issued exaggerated claims about territorial gains; AI‑generated visual material provides a seemingly concrete backdrop that makes those assertions appear more credible. By projecting an image of relentless Russian advance, the Kremlin seeks to bolster its narrative that it retains the battlefield initiative, thereby influencing foreign policymakers, journalists, and potential supporters. As of August 31, ISW noted that “most settlements claimed by Russian sources using suspected or assessed AI‑altered footage remained under Ukrainian control,” underscoring the disconnect between the manufactured visuals and the reality on the ground.
Future Outlook and the Need for Vigilant Verification
Looking ahead, ISW anticipates an expansion of AI‑generated content as video‑generation tools become cheaper and more widely available. The think‑tank warns that Russian actors could eventually employ AI‑generated satellite imagery to fabricate evidence of military movements or territorial control, adding another layer of complexity to open‑source intelligence efforts. Despite advances in detection technology, visual anomalies—such as odd proportions, lighting inconsistencies, and impossible object interactions—remain common enough to serve as reliable indicators for investigators. ISW concluded that “open‑source researchers will need to continually refine methods for distinguishing authentic battlefield footage from AI‑generated material,” highlighting the ongoing cat‑and‑mouse game between propagandists and fact‑checkers in the digital age.
https://www.kyivpost.com/post/84583