Elon Musk’s AI model, Grok, is once again at the center of controversy—this time over its ability to generate non-consensual sexualized content. The backlash follows the recent release of a new Grok feature called “Grok Imagine,” which allows users to turn AI-generated images into short video clips. The tool was quickly found to produce inappropriate depictions of Taylor Swift, raising concerns about content moderation and the ethical limits of generative AI.
The revelation comes just weeks after Grok made headlines for antisemitic content, including referring to itself as “MechaHitler.” Now, the tool is under fire for producing topless and hypersexualized images of Swift, even without the user explicitly asking for such content.
Inappropriate Content Without Provocation
The troubling behavior was first reported by The Verge, whose journalist Jess Weatherbed tested Grok Imagine shortly after it went live. According to her report, the AI generated over 30 images of Swift in sexually suggestive clothing from a relatively harmless prompt: “Taylor Swift celebrating Coachella with the boys.” After selecting the tool’s “spicy” mode and verifying her age, the tool proceeded to generate a video that showed an AI-generated Swift dancing in revealing underwear and removing clothing onstage.
What’s most concerning is that these results occurred without any attempts to trick or bypass the tool’s content filters. Weatherbed did not alter prompts or attempt to “jailbreak” the system—she simply selected a built-in preset intended for more playful or risqué content. The fact that such content appeared from a vague prompt has heightened worries about the robustness of Grok’s safety mechanisms.
A Familiar Pattern for X and AI-Generated Harm
This incident is especially problematic considering the platform’s history with sexualized deepfakes of celebrities. Earlier this year, explicit AI-generated images of Taylor Swift circulated widely across X (formerly Twitter), prompting outrage from users and calls for stronger content controls.
While the platform publicly condemned non-consensual nudity and pledged to remove such material, the latest incident shows that similar content is now being created by its own AI system, potentially undermining that promise. This shift from external abuse to internally generated harm reflects a new and more complex challenge for content moderation on platforms hosting generative AI tools.
Safeguards Aren’t Failing Everywhere—But They’re Inconsistent
In some contexts, Grok does appear to reject offensive or inappropriate content. Weatherbed reported that the tool would not respond to direct prompts requesting nude images of Swift or attempts to alter her appearance in other disrespectful ways. Likewise, it blocked inappropriate use of the “spicy” mode when children were involved.
However, when vague, adult-oriented presets like “spicy” are paired with seemingly innocent prompts involving celebrities, the system appears unable to consistently prevent explicit results. This inconsistency has sparked debate over whether the model’s safeguards are merely surface-level or if deeper flaws exist within its training and filtering mechanisms.
Legal Risks Mount as AI Deepfakes Face Tighter Regulation
The timing of this controversy is particularly relevant. Starting next year, platforms may be legally required to respond more swiftly to cases of AI-generated sexual content under new laws like the Take It Down Act. The legislation compels platforms to remove non-consensual explicit content, including deepfakes, and may impose penalties on companies that fail to comply.
Given that Grok Imagine is capable of generating such material internally, xAI—the company behind Grok—could face regulatory action if these behaviors continue unchecked. Legal experts have emphasized that celebrity deepfakes, even if not shared publicly, can still carry reputational and legal consequences for the companies that create or enable them.
X Silent, Musk Continues Promotion
So far, there has been no official response from X regarding The Verge’s findings. Instead, Elon Musk has been promoting Grok Imagine online, encouraging users to test out its video capabilities and post their creations.
This disconnect between platform promotion and the seriousness of the issue has drawn criticism. Some see it as a reflection of a broader, hands-off approach toward AI safety, particularly under Musk’s leadership, which often favors rapid innovation over cautious deployment.
