Does an authentic photo show an agent from Immigration and Customs Enforcement restraining a distressed little girl? No, that's not true: This image is AI generated. Hive, an AI-generated image detection service determined with 97.4 percent confidence the image was "highly likely" to be AI generated. Another service, Sightengine, returned the same reading with a 97 percent confidence score.
The image appeared in a post (archived here) on X on January 27, 2025. The post was captioned:
This is what conservatives voted for
This is how the post appeared at the time of writing:
(Source: X screenshot taken on Tue Feb 04 15:39:46 2025 UTC)
Lead Stories used the Hive Moderation browser extension to test the image posted on X (pictured below). The input was determined with 97.4 percent confidence "likely to be AI-generated." The tool also determined with 63 percent confidence that the image was made by the AI image generator Stable Diffusion.
(Source: X screenshot with Hive extension taken on Tue Feb 04 15:45:46 2025 UTC)
A second analysis was done with sightengine.com (pictured below). The image was again determined to be "likely AI-generated," with a 97 percent confidence score. Stable Diffusion was again determined to be the source of the image, but with a 92 percent confidence score.
(Source: sightengine.com screenshot taken on Tue Feb 04 15:56:46 2025 UTC)
Several fact checking organizations have published articles on this image that was first posted on X (archived here) on January 24, 2025, by the account @LiveOnTheChat. A January 28, 2025, Snopes article (archived here) reported they reached out to @LiveOnTheChat -- who not only confirmed that he had written the prompt to generate the image, but that he had used the xAI chatbot tool Grok to do it. He provided Snopes with the four images Grok generated with this prompt:
generate an image of Police ICE agents aggressively dragging latino children crying out of a 2nd grade classroom
The AI generated image detection platforms Hive and Sightengine do not include Grok on their lists of potential image sources, so Grok could not have been named as the potential source in their analysis. Images generated by Grok do have a visible but discreet watermark in the lower right corner. In the case of the image posted by @LiveOnTheChat, the right side of the image contained some glaring flaws and was cropped off along with the identifying watermark. There is no doubt that @LiveOnTheChat was the source of the AI generated image, having both the prompt and the original uncropped image.
At the time this was written, other fact checking agencies like The Dispatch and AFP Fact Check had reviewed this claim.
Additional Lead Stories articles on claims involving ICE agents can be found here.
