> At a Glance
> – Elon Musk’s Grok chatbot is being used to create ultra-explicit sexual videos on its stand-alone website and app.
> – A cache of 1,200 Imagine URLs reviewed by News Of Fort Worth reveals blood-soaked, weapon-involved, and celebrity deepfake clips-~10 % flagged as possible CSAM.
> – The material is far more graphic than public X posts and may include sexualized imagery of apparent minors.
> – Why it matters: AI-generated CSAM is illegal in many countries, and regulators in Europe have already received ~70 reported links for investigation.
Grok’s video generator, available only outside X, is quietly hosting some of the most violent and sexualized AI content yet seen. A new audit shows users have weaponized the tool to produce photorealistic abuse scenes and deepfakes with minimal safety friction.
Inside the Hidden Video Archive
Around 800 of the archived links contain videos or stills made by Grok since last August. Paul Bouchaud, lead researcher at AI Forensics, told News Of Fort Worth the stash is “overwhelmingly sexual content”-spanning manga, hentai and photorealistic porn, many with audio.
- Full nudity and penetrative sex acts rendered in HD video
- Blood-covered characters and knives inserted into genitalia
- Celebrities and news anchors spliced into non-consensual nude scenes
- Fake Netflix posters showing AI Princess Diana in a threesome
Because Imagine links are not public by default, most uploads stay private unless users share the URL. Google indexing and a deepfake porn forum, however, exposed hundreds of examples.
Child-Safety Red Flags
Bouchaud estimates that slightly less than 10 % of the 800 samples depict what appears to be child sexual abuse material-some in cartoon form, others “very young-appearing women” in photorealistic videos.
AI Forensics has already submitted ~70 URLs to European regulators. French prosecutors recently opened an investigation after lawmakers complained about “stripped” images on X; the new video evidence could widen that probe.
| Content Type | Approx. Share | Notes |
|---|---|---|
| Manga / Hentai | Majority | Explicit, often violent |
| Photorealistic porn | Substantial | Includes celebrity deepfakes |
| Suspected CSAM | <10 % | Mix of anime and lifelike imagery |
Key Takeaways
- Grok’s Imagine model produces far more extreme content than its public X output.
- Roughly 1,200 links archived since August point to a hidden ecosystem of AI porn, some bordering on illegality.
- Regulators in Europe have been handed ~70 cases for possible CSAM violations.
The discovery spotlights how quickly generative AI can outpace platform safeguards-and how private URLs can still find their way onto the open web.
