28 August 2026
Lawsuit claims Grok trained on child abuse material and generated more
First reported
Ars Technica ran this on , a day before the next source picked it up.
- A woman identified as Jane Doe filed a lawsuit alleging that images of her childhood abuse, documented by child protection organizations, were used to train xAI's Grok image generator.
- The complaint claims Grok generated new abusive images based on her original photos, and that xAI then used those AI-generated images to further train Grok, according to Doe's legal filing.
- Doe seeks class action status to represent all abuse survivors whose images were used this way, and asks the court to force xAI to destroy generated images and prevent Grok from creating sexualized content.
- xAI's terms of service treat its own outputs as training data by default, and do not specify whether abuse material is excluded from that pipeline, the lawsuit states.
How it was covered
The AlgorithmMIT Technology Review
A lawsuit claims that victims' images of child sexual abuse material entered Grok's training data, and a federal judge noted that AI has outpaced child-abuse laws.
Reported by Ars Technica