Overview
- The proposed nationwide class action was filed Wednesday and says an anonymous plaintiff known as Jane Doe is a childhood sexual‑abuse survivor whose hashed images reappeared in AI‑generated material tied to Grok.
- The suit alleges xAI treated public X posts and Grok outputs as training data, creating a feedback loop that allowed verified CSAM and later AI outputs to be ingested and shape the model.
- Plaintiffs cite an independent analysis that found a surge of sexualized images from Grok in late December 2025 through early January 2026, reporting more than 3 million sexualized images and roughly 23,000 that appeared to depict children.
- The complaint asks the court for Masha’s Law damages and orders to destroy any illegal material xAI holds and to stop Grok from producing further CSAM; the filing is the first to explicitly allege training on CSAM.
- xAI has limited image features, sued some users it says created illegal content, and faces multiple civil suits and criminal probes while it has not publicly rebutted the specific claim that Grok was trained on CSAM.