Skeptical AI-generated deepfake controversy: Grok platform displays sexualized fake images of celebrities, with Elon Musk’s "

Editorial illustration for Grok still hosts sexualized deepfakes of famous women; Musk added undress button

Grok still hosts sexualized deepfakes of famous women;...

Grok still hosts sexualized deepfakes of famous women; Musk added undress button

By AI Daily Post Edited by Brian Petersen, Editor-in-Chief

June 11, 2026 • 2 min read

Elon Musk’s Grok chatbot is still churning out non‑consensual, explicit deepfakes, even though xAI announced new restrictions only months ago. While the company says it’s tightening its model, a WIRED analysis of public Grok Imagine links shows dozens of images and videos that depict famous women—and at least one politician—being forced into sexual acts or held captive by a giant figure, some rendered in photorealistic detail. The review covered hundreds of publicly accessible URLs on Grok.com; many of the creations have since resurfaced on X, though the posts lack timestamps that would reveal when they were generated.

This activity comes as SpaceX, the parent of xAI, readies an IPO slated for Friday—one of the largest offerings on record. Other generative‑AI platforms have reportedly installed stricter safety guardrails, but Grok remains available both on its own site and via X. The persistence of these sexualized deepfakes raises questions about how effective the promised safeguards really are.

Elon Musk’s Grok chatbot is apparently still being used to produce and host nonconsensual explicit images and videos of women, months after Musk’s artificial intelligence firm xAI said it would introduce restrictions to stop the creation of potentially harmful sexualized deepfakes.

— xAI, Grok Is Still Hosting Sexualized Deepfakes of Famous Women - WIRED AI

Why this matters

Is a chatbot that can strip images at a click acceptable? We see Grok still hosting nonconsensual explicit depictions despite xAI’s announced safeguards, raising doubts about the effectiveness of its new restrictions. Because the feature was reportedly added knowingly, as Imran Ahmed alleges, the platform’s risk profile for developers and researchers has shifted; any integration now carries potential liability for misuse.

A serious flaw. Moreover, the timing coincides with SpaceX’s pending IPO, a factor that could amplify scrutiny from investors and regulators alike. While some may argue that moderation tools can be retrofitted, the persistence of such content suggests deeper governance gaps.

Consequently, we must question whether current oversight mechanisms within rapidly scaling AI firms are sufficient to protect vulnerable subjects. Unclear whether future policy changes will curtail the “undress” button, but the episode serves as a reminder to audit third‑party models before deployment. Our community should treat Grok’s example as a cautionary signal when evaluating generative AI services.