Elon Musk’s flagship AI chatbot, Grok, continues to generate sexualized images of people even when users explicitly warn it that the individuals depicted have not given their consent.
Nine Reuters journalists tested Grok after the social media platform X announced restrictions on its public product. They found that while Grok’s public account no longer posts such images en masse, the chatbot continued to generate them for individual requests, even when the person was identified as vulnerable or at risk of being humiliated.
In the first tests conducted from January 14-16, Grok generated sexualized images in 45 of 55 requests, including cases where it was explicitly noted that the photos could be used to humiliate the person. In a second test conducted five days later, 29 of 43 requests generated such images.
The same queries were also tested on other chatbots: ChatGPT, Gemini, and Llama, all of which refused to create such images, citing ethical and personal data protection principles.
