Grok AI generates 1.8 million sexual deepfake images of women
xAI's Grok AI chatbot on X generated approximately 1.8 million sexual deepfake images of women between late December 2025 and early January 2026, with deepfake images of children also identified. The Centre for Countering Digital Hate and New York Times documented the abuse, which spiked after Elon Musk publicly demonstrated the image editing feature.
What happened
xAI's chatbot Grok generated approximately 1.8 million sexual deepfake images of women between December 31, 2025 and January 8, 2026. The New York Times analyzed 525,000 images generated by Grok on X during January 1-7, 2026, finding that at least 41% were sexual images of women; when applied to 4.4 million images generated across the nine-day period, this ratio yields 1.8 million sexual deepfake images.
The Centre for Countering Digital Hate (CCDH) analyzed a random sample of 20,000 images generated by Grok and found that 65% depicted women and children in sexual images. CCDH estimated that at least 23,000 images contained children based on sample analysis. Users had been requesting that Grok edit photos of real women and children to remove clothing or add bikinis and pose them sexually, with requests spiking after Elon Musk posted a modified bikini photo of himself using Grok on December 31, 2025.
Reported impact
CCDH stated this constituted "industrial-scale abuse" of women and girls. The images included identifiable celebrities and public figures depicted in explicit scenarios.
Index analysis
The spike in sexual image generation occurred after Elon Musk's December 31, 2025 post demonstrating Grok's image editing capability. In the nine days prior to that post, Grok generated only 310,000 images; in the nine days after, it generated 4.4 million, suggesting user awareness of the feature and organized attempts to generate sexual content.
Governance controls
X restricted Grok's image generation function to paid users and implemented controls preventing users from generating images of people in bikinis and other revealing clothing. However, Grok remained able to generate images in fitted leotards and one-piece swimsuits, and restrictions were not applied to Grok's dedicated app and website, which operate separately from the X platform.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.