Earlier today, the Federal Bureau of Investigation published on the social platform X two grainy, low-quality surveillance photographs depicting an unidentified individual considered a person of interest in the recent shooting involving right-wing political activist Charlie Kirk. Almost immediately after these official images began circulating online, a wave of users responded by generating and sharing so-called “enhanced” versions of the blurry snapshots, produced through a variety of artificial intelligence systems. These AI-generated reinterpretations quickly transformed what were once vague, pixelated outlines into sharp, seemingly high-definition portraits, presenting the illusion of clarity and precision that was absent from the source material. However, contrary to the impression such images may give, artificial intelligence does not possess the ability to retrieve hidden truths embedded within poor-quality visuals; rather, it constructs educated guesses through pattern recognition and probabilistic inference. In other words, what appears to be newly revealed detail is not an authentic discovery but rather a speculative reconstruction of what might exist — a process that is inherently fallible and prone to fabricating elements that were never present in reality.
Dozens of these so-called refined versions proliferated beneath the FBI’s original post. Some were almost certainly produced using the platform’s proprietary Grok chatbot, while others bore the hallmarks of external applications such as ChatGPT and similar AI-based tools. The resulting images varied greatly in plausibility. A few might superficially resemble credible attempts to sharpen the visual fidelity of the FBI’s photographs, yet many clearly deviated from believability. In one striking example, an AI-driven “text-to-image reconstruction” depicted the subject wearing attire discrepant from the original—an entirely different shirt—while also imagining strikingly exaggerated physical traits, such as a dramatically pronounced, “heroic” jawline more suited to an internet meme than a legitimate forensic enhancement. Although the stated purpose of these creations was ostensibly to aid in the identification of a suspect, their viral appeal also ensured they became a convenient vehicle for users to capture attention, garner likes, and encourage reposts across the digital ecosystem.
Despite their alluring clarity and the ease with which they spread, these artificially augmented images are no more reliable—indeed, often far less reliable—than the officially released photographs from the Bureau. History offers sobering evidence of how misleading such AI techniques can be. In one widely discussed incident, an attempt to “depixelate” a low-resolution image of former President Barack Obama resulted in the astonishing error of transforming him into a white man, entirely erasing the subject’s most defining physical feature: his race. In another case, AI interventions inserted a non-existent bulge onto the top of former President Donald Trump’s head, a detail that never existed in the original source but was conjured solely because the algorithm inferred it might belong there. Such examples underscore how AI completes images not by reconstructing factual content, but by improvising based on vast datasets of unrelated images, thereby generating plausible yet fundamentally speculative additions. While this extrapolative capacity can serve harmless or even productive purposes—say, in artistic renderings or as a tool for entertainment—it becomes dangerously misleading when relied upon as a form of evidence in matters as weighty as a criminal investigation or manhunt.
Thus, although crowds of online participants may perceive their AI modifications as contributing to the pursuit of justice, the truth is that these outputs should be treated with severe caution. At best, they represent imaginative guesses about what the person in question might resemble; at worst, they propagate misinformation that muddies public understanding and risks misidentifying innocent individuals. The FBI’s officially released photographs, blurry though they are, remain the only authentic visual records of the individual of interest, and they alone can be considered legitimate reference points. Below, for purposes of clarity, the Bureau’s original post has been reproduced, followed by several examples of the so-called AI “enhancements” that have emerged across social media.
Sourse: https://www.theverge.com/news/776793/charlie-kirk-fbi-person-of-interest-photos-shooter
