Images depicting President Donald Trump kissing White House aide Natalie Harp on a golf course circulated widely on social media last week, attracting millions of views before being identified as artificial intelligence-generated fabrications. The images, which included an unusually clear version and a blurry variant, first emerged online on the evening of Monday, September 14, according to reverse-image searches conducted by multiple sources.
Analysis by AI detection tools developed by OpenAI and Google confirmed that the images were generated using generative AI technologies. Among the most noticeable anomalies in the photos was a distorted human hand with six fingers, a common flaw in AI-generated imagery. The White House responded to inquiries by dismissing the images as deliberate misinformation intended as a political attack, attributing their spread to opponents lacking substantive critiques.
Despite obvious signs of manipulation, the images were shared extensively. One social media user, posting under the name Mykhailo Rohoza, circulated the clearer photo and commented on the contrasting impact of such images in today’s political environment versus three decades ago. Rohoza later stated he did not create the original image but used AI tools to enhance a blurry version he found online. He acknowledged that the image’s abnormalities, including the extra fingers, were present before his modification.
The images also found their way into the content strategy of some online influencers. Left-wing commentator Keith Edwards used the clearer version of the photo as a YouTube video thumbnail, although he did not include or address the image’s contents within the video itself. Edwards’ team reportedly sourced the image online but did not verify its authenticity or the identity of the person pictured.
Experts specializing in AI-generated media highlighted the rapid spread and difficulty of moderating such manipulated content on social platforms. Dartmouth College’s Hany Farid described the situation as “death by a thousand cuts,” underscoring the erosion of trust in photographic evidence as AI-generated fakes become more sophisticated and widespread.
Journalists and fact-checkers have emphasized the complex and time-consuming nature of verifying image authenticity in the current technological landscape. BBC journalist Shayan Sardarizadeh conducted an extensive analysis, confirming through multiple AI detection tools that both the clear and blurry images were artificially generated. He discovered that the clearer version was flagged by OpenAI’s detection system, indicating creation via their generative tools, while the blurry image initially evaded detection but was eventually identified using Google’s AI checks.
The source and original poster of the images remain unidentified, and it is unclear which version appeared first. The emergence and viral spread of these AI-manipulated photos illustrate the increasing challenges faced by journalists, platforms, and the public in discerning truth in an era dominated by artificial intelligence.