Elon Musk’s artificial intelligence company, xAI, is facing a significant lawsuit accusing it of training Grok models on child sexual abuse material (CSAM). The complaint, filed on Wednesday, alleges that the company utilized both real images and AI-generated content to refine Grok’s image and video generation capabilities.
The Jane Doe Complaint
The plaintiff, identified as Jane Doe, was a victim of abuse as a preschooler in the early 2000s. Her images have been hashed by organizations such as the National Center for Missing and Exploited Children (NCMEC). According to the lawsuit, the Canadian Centre for Child Protection (CCCP) notified her that it had identified AI-generated CSAM on xAI depicting her, leading to severe re-traumatization.
Training Mechanisms and Data Recycling
The lawsuit claims that xAI incorporated material included in the NCMEC Hash List into its training dataset. Furthermore, it highlights that xAI’s terms treat public X posts and Grok’s own outputs as training data by default. This suggests that if Grok generates illicit material, that material is fed directly back into the training pipeline, further shaping the model.
Doe’s complaint argues that completely removing the influence of a training example from a pre-trained model is technically difficult. Consequently, any CSAM ingested before a takedown likely continues to influence the model’s outputs even after the original images are removed from public view.
Legal Demands and Implications
The proposed class action accuses xAI of violating federal child pornography laws and Masha’s Law, which grant survivors the right to sue over production, possession, and distribution. The plaintiff is seeking:
- Monetary damages for every victim whose childhood images were used.
- The destruction of all Grok-generated CSAM stored on xAI’s servers.
- A permanent block on Grok generating any sexualized outputs, including non-consensual intimate imagery (NCII).
X did not respond to requests for comment regarding the allegations.
