AI Upscales Misrepresent FBI Person-of-Interest Photos

Earlier on Sep 11, the FBI posted two blurry photos on X of a person of interest in the fatal shooting of Charlie Kirk. Almost immediately, users responded under the FBI post with AI-“enhanced” upscales — some generated by X’s Grok bot, others with tools like ChatGPT — that turned pixelated surveillance shots into sharp, high-resolution portraits. Those AI variations ranged from plausible attempts to obviously invented details (one “textual rendering” added a different shirt and a pronounced “Gigachad-level” chin), and many were shared more as attention-grabbing content than reliable leads.

The Verge notes that generative upscalers don’t reveal new facts; they infer plausible features to fill missing pixels and have a documented history of fabricating details — examples include depixelating a low-res Obama image into a different-looking person and adding a nonexistent lump to another public figure’s head. Because these tools extrapolate rather than verify, the story warns they shouldn’t be treated as hard evidence during an active investigation or manhunt.