AI Bot’s Controversial Editing Feature Raises Consent Concerns
xAI’s Grok has landed in serious trouble. A recently rolled-out feature lets users edit photos through the bot on the X platform — no notification to the person in the image, no consent required. What’s made it worse is that Grok apparently has few meaningful safeguards against sexually suggestive content. In recent days, users have been uploading altered images of women and children in compromising scenarios, and the volume of it has been genuinely disturbing.
It seems to have started with adult-content creators who wanted playful, risqué edits of their own photos. That didn’t last long. Soon enough, people were running the same manipulations on photos of others — mostly women — without asking. Outlets including Metro and PetaPixel have reported a sharp rise in deepfake creations tied directly to Grok’s editing feature.
One particularly troubling example was a since-removed post showing two young girls posed in revealing clothing. After it surfaced, a user prompted Grok to apologise for the “AI image of two young girls in sexualized attire,” calling it a serious failure of safeguards that could breach both xAI’s own policies and US law. The legal picture around AI-generated imagery is still murky in many respects, but producing realistic, explicit depictions of identifiable minors carries real legal risk. In a strange exchange, Grok told users to report such content to the FBI as potential Child Sexual Abuse Material (CSAM), and said it would address the “lapses in safeguards” as a matter of urgency.
Those apologies read more like canned responses than any real admission of fault. When journalists sought comment, xAI gave Reuters a one-line reply — “Legacy Media Lies” — and did not respond to The Verge’s inquiry before publication.
Some of the momentum behind this came from Elon Musk himself. After he jokingly asked Grok to swap actor Ben Affleck’s face in a popular meme for one of him wearing a bikini, things escalated quickly. Images appeared of North Korean leader Kim Jong Un in a multicolored bikini, with US President Donald Trump beside him in matching swimwear — the kind of absurdist content that quickly spawned jokes about nuclear conflict. A photo of British politician Priti Patel was edited into a bikini image just days ago, adding to the pile of surreal alterations circulating on the platform.
A lot of this is clearly meant as a joke — Musk himself reposted a toaster in a bikini — but plenty of it crosses into territory that’s harder to laugh off. Users have been directing Grok toward explicitly suggestive imagery, requesting skimpy bikinis or the outright removal of clothing. Grok has generally gone along with these requests, though as of The Verge‘s observation it has stopped short of producing fully uncensored nudity.
Musk’s AI products are pitched as bold and unfiltered, with fewer restrictions than their competitors. xAI’s AI companion, Ani, reportedly engaged in flirtatious exchanges with a Verge reporter, and a separate test found that Grok could generate topless deepfakes of Taylor Swift, despite a stated policy against exactly that kind of content. Rivals like Google’s Veo and OpenAI’s Sora have put stricter guardrails around non-consensual sexual content. Meanwhile, cybersecurity firm DeepStrike has reported rapid growth in deepfake imagery, much of it involving non-consensual sexualised depictions. A recent survey of US students found that 40 percent were aware of deepfakes involving people they knew, and 15 percent had come across non-consensual explicit content.
When pressed on why it was turning images of women into bikini photos, Grok pushed back on the framing, saying: “These are AI creations based on requests, not real photo edits without consent.”
That answer has satisfied almost no one. The gap between what Grok is doing and what it claims to be doing is wide, and the question of who bears responsibility — xAI, the platform, or the users — is one the company seems content to leave unanswered for now.
