xAI Takes Grok User to Court Over Alleged Abusive Deepfakes

Disclaimer: This article discusses nonconsensual sexual imagery and child sexual abuse material. 

xAI is suing a Grok user it accuses of repeatedly changing rejected prompts until the chatbot produced sexual deepfakes of adults and children.

Elon Musk’s AI company filed the case against South Carolina resident Terry Wayne Harwood in federal court in Texas. Reuters reported that Harwood allegedly bypassed safety controls and turned ordinary photographs into explicit images without consent.

People who have never used Grok can still be pulled into the abuse. A stranger can copy a school portrait or social post into an AI service without the subject’s knowledge. By the time the person finds out, the altered image may be circulating. 

Repeated prompts after the chatbot refused

Harwood opened two accounts in December 2025 and used them for about two months, according to the federal complaint. He uploaded nonsexual photographs of adults and minors before requesting explicit images and videos.

Safety filters blocked numerous requests, but Harwood kept changing his prompts and trying again, xAI says. One incident involved a fully clothed girl who appeared to be about 10 or 11 years old. Grok rejected the first request, then made further attempts with altered wording.

Some output reportedly involved minors and constituted child sexual abuse material. Company staff eventually detected the activity and reported it to authorities.

Harwood now faces eight felony charges related to the sexual exploitation of minors in South Carolina. xAI says at least some images connected to the criminal case were generated or altered through its service. 

Neither the charges nor the civil claims have been proven in court.

Contract claim and requested damages

Company lawyers filed a breach-of-contract claim separate from the criminal case. Harwood accepted user rules banning the sexualization of real people without consent. Additional terms prohibit child exploitation and attempts to evade safety controls.

Lawyers are seeking unspecified damages and a permanent order blocking Harwood from opening another account or using the chatbot. They also want him to reimburse the company if victims file claims over the images.

Under an indemnification clause in the user agreement, Harwood could be required to cover legal fees and other costs linked to future lawsuits. A ruling in xAI’s favor would place part of its financial exposure on the person accused of misusing the service.

Ordinary photos and the limits of user accountability

Ironically, xAI is also being sued by teenagers who say another person used Grok to create sexual images from their photos. Their case challenges the same safeguards the company says Harwood repeatedly defeated. People whose images are copied have no control over how the service responds or whether it blocks the request.

Harwood’s case could let xAI pursue an abusive user and recover related costs. Financial recovery starts after the damage. Readers need AI systems to reject the request before a nonconsensual sexual image is created.

Reducing exposure to exploitation

Harwood’s case exposes how little control people may have once someone else obtains their photo.

Personal precautions cannot stop another user from misusing an AI tool, but limiting access to high-resolution images can reduce how easily they are copied. Parents can also check whether schools, sports teams or community groups maintain searchable galleries and ask for tighter viewing controls.

Avoid posting images that reveal a child’s full name or school. Location details and daily routines can make it easier to connect a photo to the person shown.

If a suspected deepfake appears, preserve evidence before reporting it. Save the web address and take screenshots. Record the account name and when the content appeared, then avoid sharing it further.

Report the material under the platform’s rules for nonconsensual intimate imagery, impersonation or child sexual exploitation. Cases involving minors, threats or extortion may also require help from law enforcement, an attorney or a victim-support organization. Early documentation can improve the chances of removing the file and tracing the source account before more copies spread.

These measures cannot eliminate the risk. They can, however, limit unnecessary exposure, preserve information that may help identify the person responsible, and make it easier to seek removal or legal assistance.

Grok’s reported role in US operations tied to Iran is turning an xAI environmental lawsuit into a national security flashpoint.