(Credit: Thomas Fuller/SOPA Images/LightRocket via Getty Images)
Elon Musk's Grok AI has been found responding to requests for explicit images of women on X.
As flagged by Kolina Koltai, a researcher at Bellingcat, X users have been asking Grok AI to undress women in comments below their posts, 404Media reports. While the chatbot rejects prompts for completely nude images, it does fulfill "remove her clothes" requests with AI-generated images of women in bikinis or lingerie. Grok's responses are public and appear as replies to the original prompts.
When pointed to one such post and asked about its guardrails against requests for non-consensual, explicit content, Grok responded with an apology and said, "This incident highlights a gap in our safeguards, which failed to block a harmful prompt, violating our ethical standards on consent and privacy. We recognize the need for stronger protections and are actively working to enhance our safety mechanisms, including better prompt filtering and reinforcement learning… We are also reviewing our policies to ensure clearer consent protocols."
As of this writing, Grok AI continues to deliver on requests for partially naked outputs, while ChatGPT and Gemini reject them. This could be because of how the model has been trained.
We asked Grok about its policies on non-consensual explicit content. The chatbot said its "system is designed to reject or redirect" such requests, and it typically responds "with a neutral or humorous deflection, like suggesting a less spicy topic."
(Credit: Grok AI/PCMag)The report arrives as a "revenge porn" bill, called the Take It Down Act, awaits President Trump's signature. It requires social media platforms to take down non-consensual, sexually explicit content, including anything generated using AI, within 48 hours of notice.
Last year, Apple removed three apps that convert normal photographs into deepfake pornography, and San Francisco sued 16 AI-powered websites that help "undress" women.