I asked Grok's AI to undress me after X's new limits. It's still easy on the app.

In light of recent safety announcements from X, questions arise about the effectiveness of new limitations on Grok, the AI image-generation tool developed by XAI. Despite attempts to curb the generation of explicit content, Grok continues to produce sexualized images of individuals, including real people, through its mobile and web applications. After conducting a series of tests, it became clear that users can still easily request images that alter clothing through the Grok platform. The app offers functionality that allows users to modify uploaded photos with simple text prompts. For instance, when I requested Grok to 'take off my shirt,' the tool complied effortlessly. The same ease applied when I asked it to 'take off my pants' or even 'put me in underwear,' although certain requests were moderated. Grok’s capabilities extend beyond static images; it can also create videos, enabling users to generate clips of themselves undressing. During the exploration, I discovered that Grok responded positively to various requests, including more risqué prompts. While the AI refrained from displaying explicit content, it still produced suggestive images that closely approached the line. X's recent safety measures emphasized strict policies against non-consensual nudity and child exploitation. However, the updates appear to primarily target the @Grok account, leaving the standalone Grok applications less affected. Users can still generate sexualized content by using the Grok tab within the app, thus circumventing the restrictions intended to mitigate such activities. Moreover, Grok's developers claim to have implemented geo-blocking to prevent image generation in regions where it is illegal. Yet, investigations revealed that even in countries with stringent anti-pornography laws, the AI continued to produce bikini images, raising concerns about the enforcement of these measures. As the controversy unfolds, many are left questioning the actual effectiveness of X's newly established guidelines.