AI upscaling tools create fake details in FBI Kirk shooting investigation photos

Internet users are using AI tools to upscale and “enhance” blurry FBI surveillance photos of a person of interest in the Charlie Kirk shooting, but these AI-generated images are creating fictional details rather than revealing hidden information. The practice demonstrates how AI upscaling tools can mislead criminal investigations by inferring nonexistent features from low-resolution images.

Why this matters: AI upscaling has a documented history of creating false details, including past incidents where it transformed Obama into a white man and added nonexistent features to Trump’s appearance, making these “enhanced” images potentially harmful to legitimate investigations.

What happened: The FBI posted two blurry surveillance photos on X seeking information about a person of interest in the Charlie Kirk shooting.

• Users immediately responded with AI-upscaled versions created using tools like X’s Grok bot and ChatGPT.
• The AI-generated variations show dramatically different details, with some displaying obviously fabricated features like “Gigachad-level chin” and completely different clothing.
• While ostensibly meant to help identify the suspect, the images also serve as attention-grabbing content for likes and reposts.

The technical problem: AI upscaling tools don’t actually uncover hidden details in pixelated images—they extrapolate and infer what might be there based on training data.

• The technology fills in gaps by making educated guesses rather than revealing actual information.
• Different AI tools produce varying results from the same source image, highlighting their unreliability.
• These tools can be useful in certain circumstances but should never be treated as evidence in criminal investigations.

Past examples: Previous incidents demonstrate the dangerous inaccuracy of AI upscaling in real-world scenarios.

• AI tools have “depixelated” low-resolution images with completely incorrect results.
• The technology has added nonexistent physical features to recognizable public figures.
• These failures underscore why law enforcement agencies rely on original surveillance footage rather than AI enhancements.