xAI accused of training Grok using child abuse material in new lawsuit

A lawsuit filed Wednesday alleges that xAI's Grok model may have been trained using child sexual abuse material (CSAM), marking the latest in a series of allegations against the platform regarding its handling of such content.

The Complaint#

According to the complaint, a plaintiff identified as Jane Doe became a victim of repeated sexual abuse as a preschooler in the early 2000s, which was documented in images traded online. Over two decades, her images have been catalogued and hashed by child protection organizations including the National Center for Missing and Exploited Children (NCMEC) and the Canadian Centre for Child Protection (CCCP) to prevent their spread.

Doe enrolled in the Department of Justice Victim Notification System to receive alerts whenever her images are identified in new criminal investigations. She was notified by the CCCP that AI-generated versions of her likeness depicting abuse had surfaced on xAI's platform, and she discovered online discussions between offenders planning to create similar synthetic abuse material.

Training Data Concerns#

The lawsuit contends that xAI may have compounded the harm by allegedly using both the original abuse material and newly generated synthetic images to train Grok further. This approach would mean the model was exposed to decades of abuse material involving the same victim, perpetuating her exploitation in perpetuity through the machine learning pipeline.

Broader Pattern of Allegations#

This case adds to mounting scrutiny of how Grok handles harmful content. Regulators and courts have been investigating the extent of CSAM-related issues on the platform, and multiple Grok users have faced criminal charges tied to its capabilities. xAI has also faced previous lawsuits alleging it created synthetic abuse material.