An investigator showed a Wisconsin teenager a photo from her eighth-grade graduation. She was 14 in the picture. According to Bloomberg’s investigation into Grok and childhood photos, a classmate’s adult brother had used Grok to turn it into child sexual abuse material (CSAM). The images were then posted online.
She is now 18 and is known in court as Jane Doe 5. She is one of a growing group of young plaintiffs suing xAI, the Elon Musk company that builds Grok. Their cases raise one question the AI industry has not settled. When a chatbot creates illegal images of a real child, who is legally responsible: the user, the company, or both?
The lawsuits so far
The claims have spread across several states this year. The table below lists the main cases reported by Bloomberg, NPR, CyberScoop and local outlets.
| Filed | Plaintiffs | Main allegation |
|---|---|---|
| March 2026 | Three Tennessee teens (Jane Does 1-3) | Images made from yearbook and social media photos through an app that used xAI’s model |
| July 2026 (amended) | Added a Wyoming woman (Jane Doe 4) and the Wisconsin teen (Jane Doe 5) | A stepfather and a classmate’s brother used Grok on childhood photos; Stability AI added as a defendant |
| July 2026 | An Arkansas family | A Bentonville photographer allegedly used Grok on photos of clients, including their daughter |
| August 2026 | A California abuse survivor, for a nationwide class | xAI allegedly trained Grok on real abuse images of her |
Tennessee, Wyoming and Wisconsin
The first case came from three Tennessee teenagers in March. NPR reported at the time that the man involved did not use the Grok app or X directly. Instead, the complaint says he used another app that relied on xAI’s model. He allegedly made similar material of 18 other people and was reportedly arrested.
In July, the plaintiffs amended the complaint. NPR’s report on the expanded suit says the Wyoming plaintiff’s stepfather allegedly generated about 7,000 explicit images and videos from one photo taken when she was about 11. The complaint also claims xAI sent only one tip about her to the National Center for Missing and Exploited Children (NCMEC). That tip allegedly left out the generated images and the user’s IP address.
The amended suit added Stability AI, maker of Stable Diffusion, as a second defendant. Stability AI told NPR that any claim that safety is not a top priority is “categorically wrong.” CyberScoop reported that the man in Jane Doe 5’s case was arrested and charged, but much of the content remains online.
Arkansas and California
KATV reported that an Arkansas family sued in July through the Potts Law Firm. Police arrested Bentonville photographer Russell Bloodworth III on June 10 on 200 counts tied to deepfake abuse material. An affidavit says his phone held about 1,700 such images and videos.
The broadest case arrived in August. A survivor of childhood sexual abuse filed a nationwide class action in federal court in Northern California, CyberScoop reported. She claims real images of her abuse helped train Grok. Her lawyers say thousands of people could qualify as class members.
What changes for families
The harm in these cases started with ordinary photos. They include a yearbook picture, a beach photo on Instagram and a graduation snapshot. Bloomberg reports that some plaintiffs now fear their own graduations. Others describe anxiety and depression. Some victims were as young as 6, the report says.
Annika Martin, a lawyer at Lieff Cabraser Heimann and Bernstein, represents several of the victims. She told Bloomberg that the risk is not limited to adult offenders. Minors who use AI to make such images of classmates may not realize it can be a federal crime, she warned.
Steps you can take now
- Report suspected abuse images of a child to local police. Police investigations led to the arrests described in these suits.
- Use NCMEC’s free Take It Down service for explicit images of someone under 18. It creates a digital fingerprint, called a hash, on your device. The image itself is never uploaded. Participating platforms then scan for matches.
- Know that the federal Take It Down Act became enforceable in May 2026, according to a letter from state attorneys general. It covers removal requests for nonconsensual intimate images.
- Review what is public. Bloomberg notes that photos from social media accounts or school websites can be altered in seconds.
How xAI and Musk have responded
Grok was designed as a looser rival to other chatbots. It offers a “Spicy” mode that can produce adult images and video. Early this year, X users generated thousands of sexualized or “undressing” images per hour, Bloomberg reports. Some reports estimated tens of thousands of realistic sexualized images of children.
After a global backlash, X limited Grok image generation to paying users in January. Critics told Bloomberg this turned a contested feature into a paid product. In a January X post, Al Jazeera reported, Musk wrote that he was “not aware of any naked underage images generated by Grok. Literally zero.” He also argued that Grok only acts on user requests.
xAI later said it had removed Grok’s ability to make nonconsensual sexualized images of real people. The company says it has zero tolerance for child exploitation. X said in September that it reports such content to NCMEC. xAI has also sued several individuals for misusing Grok, including the Bentonville photographer. It declined to comment to Bloomberg, citing pending litigation. NPR notes the company now operates as SpaceXAI after merging with SpaceX.
The legal fight over responsibility
Several plaintiffs rely on Masha’s Law. This decades-old federal law lets victims of child sexual exploitation seek money damages. Section 230 of the Communications Decency Act may also matter. That law usually shields websites from liability for what their users post.
The two sides read the situation differently. Musk says the user is responsible, not the tool. Martin argues that Section 230 should not protect content that Grok itself generated. She told Bloomberg the users uploaded harmless, fully clothed photos, and the model created the abuse material.
The Minnesota case
xAI is also fighting a new law. Minnesota’s HF 1606 bans tools that create fake nude images of people without consent. Companies face penalties of $500,000 per unlawful use, MPR News reported. xAI sued the state in July, arguing the law violates the First Amendment. Its general counsel, James Burnham, later called the law plainly unconstitutional, per Bloomberg.
The first round went to Minnesota. U.S. District Judge Donovan Frank denied xAI’s request to pause the law, Minnesota Lawyer reported. The law took effect on August 1. The judge has not yet ruled on the constitutional claims.
Regulators have not imposed major penalties
In January, a bipartisan group of 35 state attorneys general wrote to xAI. Michigan’s attorney general said the group wants Grok to stop producing such images. It also wants existing content removed and action against the users who made it. Malaysia and Indonesia blocked Grok, and the UK opened an investigation.
Still, Bloomberg reports that no probe has led to a major fine so far. Investor Luke Rahbari of Equity Armor Investments pointed to Meta’s August settlement with state attorneys general as a benchmark. Bloomberg puts that deal at $18 billion, while NPR reported it as $17 billion. Rahbari expects a limited financial impact on Musk’s companies.
What to watch next
Three questions remain open. Courts must decide whether Section 230 covers images an AI model creates. The Minnesota ruling on free speech is still pending. Finally, nobody outside xAI has shown whether its newer safeguards work. For the plaintiffs, the images may keep resurfacing online while those questions wait.
