Grok’s ‘Spicy’ Video Setting Sparks Taylor Swift Deepfake Concerns
▼ Summary
– Grok’s “spicy” mode for its AI video tool generates uncensored celebrity deepfakes, including topless videos of Taylor Swift, unlike competitors with safeguards.
– The tool offers four video presets (“Custom,” “Normal,” “Fun,” “Spicy”), with “Spicy” enabling suggestive or partially nude content from generated images.
– Grok produced over 30 images of Taylor Swift in revealing outfits upon request, some with uncanny likeness, and easily animated them into inappropriate videos.
– The tool lacks effective age verification, requiring only a birth year input without proof, making it easy for minors to bypass restrictions.
– Despite policies banning pornographic likenesses, Grok Imagine allows creating suggestive celebrity videos, raising legal and ethical concerns.
Grok’s new AI video tool raises serious concerns with its ability to generate deepfake celebrity content, including unauthorized explicit imagery of Taylor Swift, despite existing safeguards in other platforms. The controversial “spicy” mode within Grok Imagine bypasses typical content restrictions, allowing users to create suggestive videos featuring recognizable public figures with minimal oversight.
During initial testing, the tool readily produced topless videos of Swift without explicit prompting, simply by selecting the “spicy” video preset after generating an image of the singer. While other AI platforms like Google’s Veo and OpenAI’s Sora enforce strict filters against deepfakes and NSFW material, Grok’s system appears to lack meaningful barriers. A basic age verification step, easily circumvented, was the only checkpoint encountered before generating questionable content.
The process begins with text-to-image generation, where prompts like “Taylor Swift celebrating Coachella with the boys” yielded dozens of options, some already featuring revealing outfits. Converting these into video clips took seconds, choosing the “spicy” option often resulted in Swift stripping down to minimal clothing, dancing provocatively in front of AI-generated crowds. Though the likeness wasn’t flawless, the resemblance remained unmistakable.
Disturbingly, while the tool blocked outright nudity in still images, the video function circumvented this restriction, effectively enabling partial undressing through animation. Though the “spicy” setting doesn’t guarantee nudity every time, multiple tests showed it frequently defaulting to risqué outcomes. The system also generated photorealistic images of children but, fortunately, refused to animate them inappropriately, though the questionable preset remained accessible.
This lax approach contradicts xAI’s own acceptable use policy, which prohibits pornographic depictions of real people. Yet Grok Imagine provides tools that make such content easily producible, raising legal and ethical red flags. With over 34 million images already generated in just days, and Elon Musk noting its rapid adoption, the potential for misuse is alarming. Without stronger safeguards, the tool risks becoming a vector for harassment, misinformation, and unauthorized celebrity exploitation.
(Source: The Verge)
