Elon Musk’s X platform has implemented new limitations to prevent the generation of explicit images of real people, particularly those in revealing clothing. The move follows widespread condemnation of Grok, X’s AI chatbot, for facilitating the creation of thousands of harmful and nonconsensual “undressing” images, including those depicting apparent minors.
However, despite the restrictions now in place on X itself, independent testing reveals that the standalone Grok app and website remain capable of generating sexually explicit content and “undress”-style images. Researchers at AI Forensics confirm they can still create nude imagery via Grok.com, while WIRED testing demonstrated that the system can remove clothing from images of men without restriction. The Grok app itself prompts users for their birth year before generating such content.
This inconsistency highlights a critical issue: while X appears to be cracking down on image generation within its platform, users can bypass these restrictions through the dedicated Grok interface. This suggests a fragmented enforcement strategy, allowing harmful content to proliferate outside of direct X oversight.
Investigations across multiple countries—including the US, Australia, and the UK—have already condemned X and Grok for enabling the creation of nonconsensual intimate imagery. The UK, in particular, is actively investigating the platforms.
X claims to have implemented technological measures and geoblocks to prevent the generation of revealing images in jurisdictions where it’s illegal. However, the persistence of explicit content generation on the standalone Grok platforms undermines these claims. Musk has also publicly stated that “spicy mode” allows for upper-body nudity of imaginary adults, framing this as consistent with R-rated content standards.
Generative AI systems have long been vulnerable to bypasses, with users employing “jailbreaks” to circumvent safety measures. While OpenAI and Google’s systems have similar vulnerabilities, the open nature of Grok’s interface makes it particularly susceptible to exploitation.
Users on pornography forums report mixed results, with some successfully generating nude content while others encounter stricter moderation. The ongoing cat-and-mouse game between developers and users highlights the difficulty of fully controlling AI-generated content.
“The reality is that safety measures are only as effective as the enforcement behind them. If a platform allows loopholes, malicious actors will exploit them.”
Ultimately, while X has taken steps to address the immediate outrage, the continued availability of explicit content on Grok’s standalone platforms raises serious questions about the company’s commitment to preventing abuse. The fragmented approach suggests that enforcement is selective rather than comprehensive, leaving users vulnerable to exploitation.
