X Bans Grok Bikini Edits on Real People, But AI Characters Can Still Be ‘Undressed’

X has tightened restrictions on Grok’s image-editing capabilities following widespread criticism over the creation of non-consensual sexual deepfakes. However, the update has sparked fresh debate after it became clear that while real people are now protected from certain edits, AI-generated or imaginary characters can still be sexualised.

The controversy centres around Grok, the AI chatbot developed by xAI and integrated into X. After days of backlash, the platform confirmed it has blocked Grok from editing images of real people into bikinis or other sexualised clothing. The decision was aimed at curbing the misuse of AI tools to create explicit or exploitative images without consent.

Despite the policy shift, the issue remains far from settled. Tests conducted by multiple publications indicate that Grok still allows users to create similar content when the subject is an AI-generated or fictional character. This distinction has drawn criticism from digital safety experts, who argue that the loophole weakens the effectiveness of the new safeguards.

The debate gained momentum after a public exchange involving Elon Musk, the owner of X. Responding to claims that Grok could no longer be pushed into generating sexualised images, Musk openly challenged users to test the system’s moderation. In subsequent posts, he clarified the platform’s position, stating that with NSFW settings enabled, Grok is intended to allow limited nudity for imaginary adult characters, similar to what is permitted in R-rated films in the United States.

According to Musk, this approach reflects what he described as the “de facto standard” for adult content in the US, while acknowledging that rules may differ across countries due to local laws. His comments, however, have only intensified scrutiny of how consistently those standards are being applied.

X later formalised the update through its safety channels, confirming that Grok is now technically restricted from editing images of real individuals into revealing clothing, including swimwear. The restriction applies to all users, including those on paid plans. Prompts that previously worked now often return blurred or blocked results when they involve real people.

In practice, enforcement appears uneven. Reports suggest that while direct requests are blocked, slightly modified prompts can still generate highly sexualised imagery. In several cases, users were reportedly able to produce revealing edits by altering the wording of their requests. Age verification measures were also found to be minimal, with basic self-declared age checks that could be easily bypassed.

The inconsistencies extend further. While some safeguards apply to images of women, Grok reportedly continues to generate sexualised images involving men or even non-human subjects with few restrictions. This selective enforcement has raised questions about how comprehensively the platform is addressing misuse.

X and xAI have attributed these gaps to the challenges of moderating complex user inputs and what Musk has previously described as “adversarial prompt manipulation.” Critics, however, argue that the current system still places too much responsibility on users rather than building robust protections into the technology itself.

As AI-generated content becomes more powerful and accessible, the Grok controversy highlights the growing tension between creative freedom, platform standards, and the need to prevent harm. While X’s latest move signals an attempt to respond to criticism, the continued ability to sexualise AI characters suggests the debate over ethical boundaries in AI image generation is far from over.