The artificial intelligence has made the expression “seeing is believing” increasingly outdated. A recent study warns that commercial generative tools such as Grok o Gemini They are capable of creating “in seconds” hyperrealistic but false images in which the convicted pedophile appears Jeffrey Epstein along with leaders like Donald Trump o Emmanuel Macron.
The experiment, carried out by the American anti-disinformation observatory NewsGuard, demonstrates how easily the general public can be deceived. Something that is already happening. The AFP agency has detected that users of social networks They are sharing fake images of the sex offender socializing with the mayor of New York, Zohran Mamdaniwhen he was just a child. Hoaxes visuals scattered by groups of extreme right and with millions of views on platforms like X that they are gaining ground on the truth.
Sexual predator Jeffrey Epstein. / House Committee on Oversight
Grok, they Musk
NewsGuard has applied to three imaging programs using IA that they create photos of Epstein with five leaders. Grok Imagine, developed by xAI, a company owned by Elon Muskproduced “convincing fakes in seconds.”
Gemini, the Google
Google’s Gemini refused to generate fake images featuring Trump, but had no problem doing so with other politicians such as Macron, the Israeli prime minister. Benjamin NetanyahuUkrainian president Volodymyr Zelensky and the British Prime Minister Keir Starmer.
ChatGPT, the OpenAI
According to the study, the only commercial generative AI tool that rejected requests was ChatGPTof OpenAI. The chatbot responded to NewsGuard researchers’ request by stating that it “cannot create images involving real people with sexualized depictions of minors or scenarios involving sexual abuse“.
Donald Trump y Jeffrey Epstein / Europa Press/Contacto/Epstein Estate/House Oversig
“The findings demonstrate how easily malicious actors can use AI imaging tools to generate credible fakes that go viral, and why fake images have become so common that it is difficult to distinguish authentic ones from AI-generated ones,” the NewsGuard report says.
Subscribe to continue reading
