Musk's chatbot Grok remains under fire as lawsuits pile up

Elon Musk vowed to stop the creation of AI deepfakes. His companies issued statements, introduced technical restrictions, and acted as though they had the situation under control. Yet Grok, the artificial intelligence developed by xAI, continues to create sexualized images of real people without their consent. NBC News documented this in April 2026, three months after Musk promised to address the issue.

What exactly is happening behind the walls of the X platform? And why can it not be stopped—or rather, why is there no desire to stop it?

What NBC News discovered

NBC News reporters reviewed dozens of posts published on the X platform over the past month. They found sexualized images and videos of real women that Grok had created in response to user requests. Most featured singers and actresses. Users found new ways to circumvent the system. Although Grok now rejects commands such as "put her in a bikini," it complies with others. Here are the three main techniques users are exploiting:

First, a user uploads two images at once: a real photograph of a woman and a drawing of a stick figure in a spread-legged position. They then tell Grok to make the woman "assume the same position." The result is an image of the woman with her legs spread. Second, users ask Grok to swap clothing between two different photographs, with one of the outfits being revealing. Third, photographs are turned into videos. NBC News documented a case from March 2026 in which Grok turned a single photograph of an actress into a video showing her touching her breasts. Another case from April 2026 involved a video in which the same actress is sitting with her legs spread, even though her legs were crossed in the original photograph.

None of the women in the documented images were naked, and none appeared to be minors. That is an improvement over January 2026. But it is not enough.

The beginning of deepfakes at xAI

It all erupted around the turn of 2025 and 2026. In late December, users began complaining about a flood of sexualized deepfakes depicting women and girls. At the time, Grok simply complied whenever people entered a command such as "photograph her naked" or "put her in a swimsuit." Musk's company even promoted the feature under the name "spicy mode" and publicly praised it.

On December 31, 2025, Grok admitted on the X platform that in "isolated cases," users had received images of minors in "minimal clothing." The AI also wrote that it "deeply regretted" this. But xAI did not change the product. Instead, it shifted responsibility onto users. "Anyone using Grok to make illegal content will suffer the same consequences as if they upload illegal content," Musk wrote on X.

Then came global pressure. The UK-based Internet Watch Foundation reported the presence of "criminal material" allegedly created by Grok. Researchers found that Grok was producing thousands of sexualized images per hour at the time. It was not until January 9 that X restricted image generation to paying users, and on January 14 it announced stricter measures.

Millions of images and investigations on five continents

The Center for Countering Digital Hate estimated that Grok produced more than three million sexualized images over a single eleven-day period. Independent analyst Genevieve Oh, whose deepfake research is cited by media outlets around the world, described Grok as "the largest producer of non-consensual synthetic nudity in the world". Her research suggests that Grok likely exceeds the combined output of all other similar tools. "Perverts can still use Grok to put women and girls into sexualized positions and clothing, despite the platform's claims to the contrary," said Imran Ahmed, the center's chief executive.

The scandal has attracted the attention of authorities around the world. Eight separate investigations are still ongoing. They include:

• California Attorney General's Office
• Australia's eSafety Commissioner
• Office of the Privacy Commissioner of Canada
• European Commission
• Irish Data Protection Commission
• Paris public prosecutor
• UK Ofcom
• UK Information Commissioner's Office

"The investigation in California remains fully ongoing," confirmed the office of Attorney General Rob Bonta. In addition, French authorities searched X's Paris office in February as part of the deepfake investigation. At the end of April, they plan to summon X executives, including Musk and former CEO Linda Yaccarino, to Paris for questioning. X described the search as an "abuse of authority and theater."

A Dutch court issued an order requiring Grok to stop creating images that undress adults and children. xAI argued in court that it cannot prevent all misuse of its tools and should not be punished for the actions of malicious users.

The victims do not even know about it

Behind all the statistics and court cases are real people. Stefan Turkheimer of RAINN, an organization dedicated to combating sexual violence, put it bluntly. "When these images are created and distributed, the people depicted in them usually do not even know they exist," Turkheimer said.

The National Center for Missing & Exploited Children, NCMEC, confirmed that it receives reports describing the misuse of Grok in cases where children or survivors of sexual violence may have been victimized. Musk himself wrote in January that he was not aware of a single nude image of a minor created by Grok. However, NCMEC emphasized that it had not independently examined Grok's capabilities behind the paywall.

Lawsuits are coming from all sides. Women and girls whose likenesses were altered by Grok have filed two class-action lawsuits in California. Baltimore is suing xAI for violating consumer protection law.

A platform that resists accountability

After NBC News published its findings, X issued a statement: "We strictly prohibit users from creating non-consensual explicit deepfakes and using our tools to undress real people." The company claims to have extensive safeguards, continuous monitoring, and prompt filters.

However, most of the compromising images found by NBC News disappeared from the platform only after the newsroom contacted xAI with questions. Before that, they were publicly available. And they were not difficult to find. A basic search on the X website was enough.

"Technology companies must ensure that the products they develop and make available to the public are safe by design. Sitting back and waiting until unsafe products are abused is unacceptable." A representative of the UK-based Internet Watch Foundation said.

In February 2026, xAI was acquired by another of Musk's companies, SpaceX. And in June 2026, SpaceX plans to go public at an estimated valuation of more than two trillion dollars. Legal experts warn that SpaceX will therefore likely be liable for any fines associated with Grok's conduct. Whether such fines would be "material" for a company with such a valuation is unclear. But the reputational risk is undeniable.