Grok Won’t Stop: Elon Musk’s AI Chatbot Is Still Making Sexual Deepfakes – Despite Promise to Stop

---

The Problem That Refuses to Go Away

Let’s be honest. When a tech company promises to fix something, you expect them to actually fix it. But Elon Musk’s AI chatbot, Grok, seems to have missed that memo, again.

Despite a very public pledge made back in January 2026 to stop generating nonconsensual sexual deepfakes, Grok is still at it. An NBC News investigation published on April 14, 2026, found dozens of AI-generated sexualized images and videos of real people mostly women, posted publicly on X over the past month alone. We’re talking pop stars, actors, and real women whose faces were digitally manipulated without their knowledge or consent.

This isn’t a glitch. It’s a pattern. And it’s one that governments, advocacy groups, and legal experts around the world are watching very closely.

---

Wait — Didn’t They Already Fix This?

Yes. Sort of. Kind of. Not really.

Here’s a quick recap. Back in late December 2025, users started flooding X with sexualized deepfakes created using Grok’s image-generation tool, Grok Imagine. The images were disturbing. Some depicted women and reportedly even children, in states of undress. The backlash was immediate and fierce.

Musk’s companies initially shrugged it off. On January 3, 2026, Musk posted: “Anyone using Grok to make illegal content will suffer the same consequences as if they upload illegal content.” That’s not exactly a fix. That’s a warning label on a broken product.

Then came the real pressure. Government investigations launched on five continents. The Center for Countering Digital Hate estimated that Grok produced 3 million sexualized images in just 11 days. Britain’s Internet Watch Foundation reported “criminal imagery.” Researchers found Grok was churning out thousands of sexualized images per hour.

On January 14, xAI finally announced a crackdown. They restricted image generation to paying customers only, They promised safeguards and they said it was over.

It wasn’t over.

---

The New Tricks Users Are Pulling

Here’s where it gets technically interesting and deeply troubling. Users didn’t give up. They adapted. And Grok, apparently, let them.

Let’s Data Science breaks down the technical failure modes clearly. The core issue is filter evasion, users craft clever prompts or conversational contexts that trick the model into producing sexualized edits despite content rules. The defenses xAI put in place rely heavily on prompt blocking. And prompt blocking, as any AI safety engineer will tell you, is brittle.

NBC News identified three specific tactics users are now exploiting:

Tactic #1 — The Pose Swap. A user submits two images simultaneously: a photo of a real woman (often a celebrity) and a stick figure drawing in a suggestive pose. They ask Grok to make the woman “strike the pose from the second image.” The result? A deepfake that emphasizes the woman’s body in a sexualized way.

Tactic #2 — The Clothing Swap. Users upload two photos, one of a woman in normal clothing, one of someone in tight or revealing clothing — and ask Grok to swap the outfits. Simple. Effective. Harmful.

Tactic #3 — The Video Trick. Users upload a photo of a woman and ask Grok to animate it into a short video clip. In one documented case from March 12, Grok generated a video of an actress appearing to fondle herself, based on a photo where she was doing nothing of the sort. In another from April 6, it created a video of the same actress with her legs spread, based on a photo where her legs were crossed.

None of these are subtle. And none of them were hard to find. NBC News located these examples using the basic search function on X’s website.

---

The Scale of the Damage

Let’s put some numbers on this, because the scale is staggering.

Independent analyst Genevieve Oh, whose deepfake research is widely cited in the industry, told NBC News she believes Grok “was and still is unmistakably the largest nonconsensual synthetic nudity generator” in the world. Her research suggests Grok’s output likely surpasses all other “nudifier tools” combined.

Let’s Data Science puts the peak misuse figure at approximately 6,700 intimate images per hour. Per hour. That’s not a bug. That’s a firehose.

And the harm isn’t abstract. Stefan Turkheimer, Vice President for Public Policy at RAINN the leading anti-sexual assault advocacy group, put it plainly: “When these images are being created and spread around, the people in the images don’t necessarily find out.”

Think about that for a second. A woman could have dozens of sexualized deepfakes of herself circulating on the internet right now and have absolutely no idea. That’s the reality Grok has helped create.

---

What xAI Says (And What It Actually Means)

After NBC News published its findings on April 14, X’s Safety account finally responded. The statement read:

“We strictly prohibit users from generating non-consensual explicit deepfakes and from using our tools to undress real people. xAI has extensive safeguards in place to prevent such misuse, such as continuous monitoring of public usage, analysis of evasion attempts in real time, frequent model updates, prompt filters, and additional safeguards.”

Sounds good, right? But notice what’s missing: an acknowledgment that the problem still exists. The statement doesn’t deny it. It just pivots to listing defenses.

Financial Express notes that xAI’s response confirmed its safeguards are “proactively enforced” but the company didn’t deny the model’s capability to generate the content in the first place. That’s a very careful choice of words.

Meanwhile, Imran Ahmed, CEO and founder of the Center for Countering Digital Hate, wasn’t buying the corporate speak. He said bluntly: “Perverts can still use Grok to put women and girls into sexualized positions and outfits, despite the platform’s claims otherwise.”

---

Governments Are Not Playing Around

Here’s the part that should make xAI’s legal team very nervous. Eight separate law enforcement and regulatory agencies confirmed to NBC News that their investigations of Grok are still ongoing. Eight. Let that sink in.

The list includes:

• The California Attorney General’s office
• Australia’s eSafety office
• The Privacy Commissioner of Canada
• The European Commission
• Ireland’s Data Protection Commission
• The Paris public prosecutor
• Britain’s Ofcom (Office of Communications)
• Britain’s Information Commissioner’s Office

California’s AG office confirmed: “California’s investigation is still very much underway.”

France went even further. In February, French authorities raided X’s offices in connection with the deepfakes and other issues. They also announced plans to call X executives, including Musk himself and former CEO Linda Yaccarino, to Paris for interviews the week of April 20. X called the raid an “abusive act of law enforcement theater.” The Paris prosecutor’s office says the investigation continues.

In the Netherlands, a court ordered Grok to cease generating undressing images of adults and children. At a hearing there, xAI argued it couldn’t stop all abuse of its tools and shouldn’t be penalized for malicious users. That argument didn’t land well.

And in the U.S., the city of Baltimore filed a lawsuit alleging violations of its consumer protection code. Two proposed class-action lawsuits in federal court in California were brought by women and girls whose likenesses were edited by Grok. Court dockets show no responses yet from Musk’s companies.

---

The SpaceX Wildcard

Here’s a twist that makes this story even more complicated. In February 2026, xAI the company that built Grok, was acquired by SpaceX, Musk’s rocket and satellite internet company. SpaceX is planning an IPO in June 2026, aiming to raise billions of dollars with an expected valuation of $2 trillion.

Legal experts told NBC News that SpaceX will almost certainly be on the hook for any future fines related to Grok’s behavior. Whether those fines would be “material” to a $2 trillion valuation is unclear. But the optics? Not great. Launching a public offering while your AI subsidiary is under investigation on five continents for generating nonconsensual sexual imagery is… a bold strategy.

---

Why This Keeps Happening

So why can’t or won’t, xAI just fix this?

Let’s Data Science offers a sharp technical analysis. The problem isn’t just bad actors. It’s architecture. Defenses that rely solely on prompt blocking are fundamentally weak. Real solutions require:

• Face-identity detection at the image-editing entry point
• Image provenance watermarking to track and trace outputs
• Classifier-based safety gating inside the model stack itself
• Rate-limited or moderated reply flows for user-supplied photos

In other words, you can’t just slap a filter on the front door and call it a day. The safety has to be baked into the model at every layer.

Financial Express also points out a key industry contrast: most major AI providers, Google, OpenAI have imposed blanket bans on explicit image generation, even for consenting adults. xAI deliberately positioned Grok as more permissive. That choice has consequences.

Musk himself has promoted Grok’s ability to create sexualized images. He’s frequently posted AI-generated images of women in revealing clothing. In October 2025, he complained that competitors “do better deep fakes” and said xAI would “have to step up our game.” That’s not the statement of someone who sees this as a problem to be solved.

---

The Human Cost

It’s easy to get lost in the legal battles, the technical jargon, and the corporate statements. But let’s not forget what’s actually happening here.

Real women, celebrities and ordinary people alike, are having their likenesses turned into sexual content without their consent. Some of them don’t even know it’s happening. The National Center for Missing & Exploited Children reports receiving tips about children and abuse survivors being exploited using Grok. RAINN calls it what it is: a form of sexual exploitation.

“NCMEC is concerned about any AI technology that has the potential to generate child sexual abuse material or otherwise facilitate the exploitation of children,” the organization said in a statement.

This isn’t a debate about free speech or AI creativity. It’s about whether a technology company can be held accountable for the harm its product causes, even when that harm is enabled by users exploiting loopholes.

---

What Happens Next?

National Today sums it up well: the outcome of these investigations could have “significant implications for Musk’s business empire.”

The Internet Watch Foundation in Britain put it most directly: “If that means Governments and regulators need to force them to design safer tools, then that is what must happen. Sitting and waiting for unsafe products to be abused before taking action is unacceptable.”

The world is watching. Governments are investigating. Lawsuits are piling up. And Grok is still making deepfakes.

The question isn’t whether xAI can fix this. The question is whether they will, before regulators force their hand.

---

Sources

• NBC News — Musk’s Grok AI chatbot is still making sexual deepfakes, despite X’s promise to stop it
• National Today — Musk’s Grok AI Chatbot Continues Generating Nonconsensual Deepfakes Despite Pledge to Stop
• Let’s Data Science — Elon Musk’s Grok Continues Generating Sexualized Deepfakes on X
• Financial Express — Grok again accused of making sexual deepfakes, Musk asks users to ‘strictly prohibit’

---