xAI sues Grok user accused of generating child abuse imagery

xAI sued Terry Wayne Harwood on Tuesday, accusing him of using Grok accounts to create illegal sexualized images and seeking unspecified damages. The case is a liability play for Elon Musk’s AI company as it faces claims that Grok can be used to generate child sexual abuse material and nonconsensual intimate images.

The complaint says Harwood used two xAI accounts from Dec. 8 through Feb. 18 to alter images of multiple victims, including a girl xAI said appeared to be as young as 10. xAI alleges Harwood repeatedly tried to get around Grok’s safety filters and violated both the company’s terms of service and U.S. law.

Harwood was arrested earlier this year, according to the South Carolina attorney general’s office. A spokesperson for that office said the criminal case remains pending and that Harwood has been charged over child sexual abuse material created or handled “through the use of an artificial intelligence platform.” The spokesperson was not authorized to confirm whether Grok was the platform at issue. In its civil complaint, xAI says it believes at least some images in the criminal matter were generated or altered through Harwood’s use of Grok.

A test case for AI output liability

xAI’s lawsuit asks a federal court to find that Harwood breached his contract with the company and that he, rather than xAI, bears responsibility for the outputs he allegedly obtained. The company argues in the filing that Grok is a user-directed tool and that users control prompts and resulting content.

The complaint relies heavily on xAI’s terms of service. Those rules bar users from asking Grok to undress real people, alter a person’s likeness into sexual content, portray people pornographically, or sexualize and exploit children. xAI says users agree to those terms when they sign up and that child sexual abuse material it discovers is reported to the National Center for Missing & Exploited Children.

The company says Grok’s safeguards blocked some of Harwood’s prompts, but alleges he changed wording to evade moderation. xAI did not include examples of successful prompts or describe the alleged bypass methods, a choice consistent with avoiding instructions that other users could copy. The filing also does not say Harwood received warnings from xAI before the company concluded he was barred from continued use under its terms.

xAI says Harwood should cover damages tied to alleged harm to third parties, possible lawsuits against xAI, legal costs, and reputational damage. The amount sought was not disclosed.

Class action pressure

The suit lands as xAI faces a proposed class action involving children allegedly harmed by Grok-generated sexualized images. In one recently added claim, a girl alleged that her stepfather used Grok, possibly along with other AI tools, to create about 7,000 sexualized images of her and distribute them on the dark web before dying by suicide after he was discovered.

Lawyers for that plaintiff alleged xAI failed to help police identify the user who uploaded her image. They cited a 2026 NCMEC report saying 90 percent of xAI’s CyberTipline reports lacked user information that would allow law enforcement to identify or locate suspects.

Musk previously said he had not seen examples of Grok-generated child sexual abuse material and warned users on X that illegal use of Grok would bring the same consequences as uploading illegal content. The Harwood complaint changes the company’s posture: xAI is no longer arguing only through product rules and public warnings, but through a court case aimed at making a user financially responsible for alleged illegal outputs.

Whether that argument succeeds is unresolved. The case could affect xAI’s defense against future claims if a court accepts that users are liable for both prompts and AI-generated outputs. It could also face friction with broader debates over whether AI outputs should be treated as human-created for legal purposes.

xAI did not immediately respond to a request for comment reported by Ars Technica.

This story draws on original reporting from Ars Technica.