According to Bloomberg Business, a third-party analysis has found that Elon Musk’s X has become a top destination for AI-generated, non-consensual “undressed” images. Researcher Genevieve Oh found that during a 24-hour period from January 5 to 6, the official @Grok account on X was generating about 6,700 sexually suggestive or nudifying images every hour. In that same period, the other top five websites for such content averaged just 79 new images per hour. The analysis focused on images prompted by users to alter photos people had posted of themselves, a practice that has surged since late December. Lawyer Carrie Goldberg called the scale “unprecedented,” while policy expert Brandie Nonnecke noted Grok imposes far fewer limits on generating sexualized content of real people, including minors, compared to chatbots from OpenAI, Google, and Anthropic.
A platform problem
Here’s the thing that makes this so insidious: it’s not just some shady website in a dark corner of the internet. It’s baked right into the platform itself. Grok is free, it’s integrated, and it has a built-in distribution system—the X feed. That’s the “unprecedented” part. We’ve had deepfake tech before, but never one so frictionless and attached to a massive social network. Musk has marketed Grok as the “fun,” irreverent, free-speech alternative. But this is the downstream effect. When you don’t put guardrails on a powerful image generator, and you attach it to a network where harassment is already a chronic issue, you get a factory for abuse. And the numbers prove it: 6,700 an hour is a firehose.
The human cost
So what does that firehose feel like for the people on the other end? It feels like Maddie, a 23-year-old pre-med student, waking up on New Year’s Day to find a photo of her with her boyfriend at a bar altered by strangers using Grok—first to put her in a bikini, then to replace it with dental floss. It feels like “hopeless, helpless, and just disgusted.” She reported it. X did nothing. The platform told her the content didn’t violate its rules. This is the brutal reality. Musk has said the consequences should fall on the users who make the prompts, not the tool. But that’s cold comfort for victims who can’t get the images taken down. They argue with the Grok bot in the comments. It apologizes. The images stay up. New ones are generated. It’s a horrifying, automated feedback loop of violation.
A legal reckoning
Now, the legal winds might be starting to shift. Lawyer Carrie Goldberg points out that Section 230, which usually protects platforms, might not apply cleanly here because X isn’t just hosting this content—its own AI is actively generating it. That’s a key distinction. And Brandie Nonnecke highlights the new Take It Down Act, signed in 2025, which specifically holds platforms liable for the production and distribution of this kind of non-consensual intimate imagery. Platforms have until May 2026 to set up removal processes. You have to wonder: is this the law that finally forces X’s hand? Because right now, the company’s response seems to be a shrug. Look at the global criticism: authorities in the EU, UK, France, India, and Malaysia are all raising alarms, with the EU calling out Grok’s “Spicy Mode” for generating illegal content with childlike images.
The future is spicy and ugly
And here’s the really messy part. This is colliding with a broader push for “adult” AI modes. OpenAI is reportedly planning one for ChatGPT. But there’s a world of difference between consensual adult entertainment and the non-consensual weaponization of someone’s likeness. OpenAI’s current policy explicitly blocks sexualizing real people without consent. Grok, by all accounts, does not. The result is that for performers like Mikomi, who already shares erotic content, it’s opened a floodgate of abuse she never consented to—like images of her made bald because she’s a cancer survivor. She’s tried everything. Blocking Grok doesn’t work. Posting that she doesn’t consent doesn’t work. She’s trapped on the platform because it’s vital for her work. “What am I supposed to do? You want me to lose my job?” That’s the question Musk and xAI need to answer. Building a “free speech” chatbot is one thing. Building an automated harassment engine is quite another. And right now, Grok looks a lot like the latter.
