APREX · EST. 2026
Elon Musk's AI company xAI is taking legal action against one of its own users after the individual allegedly misused the Grok chatbot to produce and distribute child sexual abuse material. The lawsuit, first reported by Reuters, targets Terry Wayne Harwood of South Carolina.
Deliberate circumvention of safety barriers
According to the lawsuit, Harwood actively worked to bypass Grok's built-in safety mechanisms before manipulating images and generating material that falls within the definition of CSAM — child sexual abuse material. xAI contends this constitutes a clear violation of the company's terms of service, The Verge reports.
Harwood was arrested back in February of this year, charged with possession and distribution of CSAM. He faces eight counts, all classified as serious criminal offences.
xAI alleges that Harwood "knowingly and intentionally used Grok to circumvent safety barriers, manipulate images without consent, and generate and distribute CSAM".
A problem growing at an alarming rate
This case is far from an isolated incident. Figures from the US National Center for Missing and Exploited Children (NCMEC) reveal a disturbing trend: the number of reports concerning AI-generated abuse material rose from 4,700 in 2023 to 67,000 in 2024 — and then surged to more than 1.5 million in 2025, according to NCMEC statistics cited in industry reports.
European organisations are documenting the same problem. The Internet Watch Foundation (IWF) uncovered more than 20,000 AI-generated images on a single dark web forum in October 2023, of which more than 3,000 were classified as criminal abuse material. During July 2024, a further 3,500 such images were uploaded to the same forum.
The industry mobilises — but the technological arms race continues
Major technology companies have not stood idle. Amazon, Anthropic, Google, Meta, Microsoft, Mistral AI, OpenAI, and Stability AI are among the players that have signed up to so-called "Safety by Design" principles, developed in collaboration with child protection organisations Thorn and All Tech Is Human. The goal is to embed child safety from the very first stage of the AI development process.
Microsoft alone reported 111,931 cases to NCMEC during 2025, while Meta contributed as many as 13.8 million reports across Facebook, Instagram, and WhatsApp in 2024.
Nevertheless, the industry itself acknowledges that it is engaged in a technological arms race against actors who continually find new ways to circumvent restrictions.
Norwegian perspective: High awareness, but challenging enforcement
Norway is not insulated from the trend. Kripos and the Norwegian Children's Ombudsman have repeatedly warned that generative AI dramatically lowers the threshold for producing abuse material — even for individuals with no technical background. Justice authorities have emphasised that AI-generated abuse images are illegal on the same basis as traditional material, even when no real children are directly involved.
The fact that an AI company is now itself taking the initiative to bring a civil lawsuit against an abuser is relatively unprecedented. It sends a signal that companies are no longer content with simply closing accounts, but are actively seeking to enforce their own terms in court — and potentially serve as a deterrent to others considering misuse.
The case against Harwood is expected to be closely watched by both the AI industry and legislators in the United States and Europe, including Norway, which is working on regulating AI-generated illegal content under the EU AI Act and existing criminal law.