What happened

Multiple AI-generated videos circulated on platforms including TikTok, X, and Telegram, allegedly showing Ukrainian soldiers surrendering en masse near the frontline city of Pokrovsk. 

The videos were crafted to appear authentic, but have telltale signs of the use of AI to generate an/or manipulate them: unnatural facial expressions, synthetic voices, and in some cases visual glitches (like stretcher-like objects “floating”).

In one of the AI-generated videos, “soldiers” speak with strange, high-pitched voices, show no real military insignia, or have strangely placed national symbols.

The videos had captions in various languages and were promoted by a network of accounts, suggesting a coordinated campaign designed to amplify Kremlin-aligned narratives globally. 

Ukrainian and independent fact-checkers confirmed the videos are deepfakes and bear a Sora watermark, and Ukrainian authorities have denied any mass surrenders, noting that actual fighting around Pokrovsk continues with Ukrainian troops holding their positions.

Why it happened

The primary motive for spreading these fake videos was to erode trust in Ukrainian leadership, frighten Ukrainian families, and influence Western public opinion and support for Ukraine. 

The use of AI enables propagandists to create hyperrealistic content at scale and in multiple languages, making disinformation more persuasive and difficult to detect. 

Lapses in platform accountability, such as slow removal of deepfake content and weak detection measures, allowed these videos to spread and reach large international audiences before being debunked or taken down. 

The incident underscores the growing weaponisation of generative AI for wartime propaganda and the persistent challenges social media companies face in policing harmful synthetic media.