The internet never forgets. The phrase serves as a reminder that anything posted online runs the risk of being accessed at will. For those who share something they probably shouldn’t have, it can be a headache. For victims of AI-generated explicit images that aren’t real, it’s much more of an ongoing nightmare.
A lawsuit filed in a California federal court on March 16, 2026, seeks to hold Elon Musk and xAI responsible for image generation containing nonconsensual explicit content created using the Grok chatbot’s tools. Three teens from Tennessee, one of whom is a minor, accuse Musk and xAI of purposely designing and marketing Grok without safeguards in place to prevent images of real people from being used to make deepfake pictures and videos containing sexual content. This includes child sexual abuse material (CSAM).
The suit, which contains 13 counts that range from intent to distribute child pornography to intentional infliction of emotional distress (IIED), seeks class-action status. This includes what it estimates to be “thousands of minors” who have been victimized by CSAM AI-generated images and videos created with Grok’s AI tools. The plaintiffs, who are listed as Jane Does, allege that Musk’s bragging about Grok’s “spicy mode” last year encouraged users to create the resulting wave of nude images of children.
X Marks a Dangerous Spot
The role of artificial intelligence (AI) in society continues to waver between what’s been promised and what it actually can accomplish. Large Language Models (LLMs) are often used in AI chatbots like ChatGPT (with 80% market share), Google Gemini, and Claude. Generative AI continues to become more sophisticated, as evidenced by how easily users can alter images and create videos. However, advances in technology can often be a double-edged sword.
Chatbots that allow users to alter images can create illegal content, and such deepfakes are often difficult to distinguish from real media. This can include intimate images that impose a victim’s face over another person’s body, a false “digital undressing” (also known as a “nudify” video), the removal of a victim’s clothing from a picture, or placing the victim in sexually suggestive positions. These deepfake images, which can contain CSAM such as nonconsensual images of minors, are then shared on social media platforms.
Legislation like the Take It Down Act, passed in 2025, makes the posting of nonconsensual intimate images (both authentic and AI-generated) illegal. Platforms hosting images and videos must implement a system that allows immediate removal of such media upon notification. Most AI companies also impose safeguards on their chatbots to prevent the dissemination of such material. These guidelines include:
- Red Teaming (pre-release testing employing outside agents attempting to circumvent the system’s safeguards)
- Training the LLM to filter out abusive and sexual content
- Pre-interference filter (rejects user requests with illegal prompts)
- Post-interference filter (recognizing and deleting CSAM after generation)
- Alignment training (human feedback as part of ongoing training)
- Hash matching (Using CSAM databases to identify and omit similar material)
- Watermarking (Identifying and marking all AI-generated images and videos)
- Zero tolerance (for any AI-generated explicit materials involving minors)
- Mandatory reporting of all CSAM materials detected
The lawsuit accuses Musk and xAI not only of eschewing these safety protocols with Grok, but of making their omission a feature rather than a bug.
The first version of xAi’s Grok, released in 2023, lacked video and audio capabilities. Subsequent versions added these features, with the latest iteration, trumpeted by xAI in October 2025, featuring “Spicy Mode,” which would allow X users to explore “edgier, more visually daring narratives.” The lawsuit includes the system prompts and safety instructions posted in November of that year, which noted that Grok would “assume good intent” by its users and that there are “no restrictions on fictional adult content with dark or violent themes.”
The complaint alleges this approach, combined with a lack of safeguards, encouraged users of Grok (often through professional tools licensed to third-party companies) to create images and videos of sexual situations and revenge porn. The Center for Countering Digital Hate estimated that between December 29, 2025, and January 8, 2026, Grok’s tools created over three million sexualized images, with around 23,000 of them including minors. This, the plaintiffs charge, was by design, allowing Musk and xAI to target a demographic seeking an AI capable of creating and altering images with or without the subject’s consent.
Really? Yearbook Photos?
The three Tennessee teens bringing the lawsuit were all victims of the same person, who gathered innocent pictures from school yearbooks, Homecoming dances, and social media accounts. The perpetrator then used Grok’s “Spicy Mode” through a third-party vendor to create highly sexualized images and videos, which they uploaded and traded on platforms like Telegram, Discord, and Mega. The suspect was arrested in late December 2025 after Jane Doe 1, alerted to the existence of the deepfakes by an anonymous Instagram account, contacted law enforcement. The subsequent investigation uncovered nonconsensual sexual images of Jane Does 2 and 3, the former of which was a minor at the time.
The suit argues that posts by Musk and xAI, along with the purposeful lack of proper safeguards, showed that Grok’s creators intended and encouraged the chatbot to be used for sexually charged creations, with no regard for nonconsensual victims. In addition to three counts involving CSAM, the complaint alleges negligence, intentional and negligent infliction of emotional distress, and design defects. With a massive number of plaintiffs possibly signing on and relief that includes requesting $150,000 per victim per violation, any jury award may have a profound effect on certain AI chatbot tools going forward.
Related Resources
- How To File a Class-Action Suit (FindLaw’s Filing a Lawsuit)
- Can Elon Musk Afford To Break the Law? (FindLaw’s Legally Weird)
- Top 5 Tips for “Revenge Porn” Cases (FindLaw’s Law and Daily Life)
