What happened: xAI's first lawsuit against its own user

On July 15, 2026, xAI filed a lawsuit in federal court in Texas against Terry Wayne Harwood, a resident of South Carolina, accusing him of having "knowingly and intentionally used Grok to circumvent safety measures, alter non-consensual images, and create and distribute child sexual abuse material (CSAM)". According to court documents reported by both Reuters and The Verge, this is one of the first cases where an AI company is actively suing its own user for misusing its own tool.

Harwood was arrested back in February 2026 and faces eight criminal charges for possession and distribution of child pornography. The lawsuit now claims that "at least some" of the images related to his criminal prosecution "were created or altered" using Grok. According to xAI, Harwood uploaded ordinary — non-sexual — photos of adults and minors into the chatbot and attempted to use AI to convert them into sexually explicit deepfakes.

From "spicy" mode to a global scandal

The case doesn't exist in a vacuum. Grok went viral last summer after xAI launched the so-called "spicy mode" — a mode that explicitly enabled the generation of sexually suggestive content. In August 2025, The Verge reporter Jess Weatherbed demonstrated that Grok Imagine easily generated nude deepfakes of Taylor Swift within seconds — without any meaningful safety barriers.

In early January 2026, the situation escalated. Grok began mass-generating sexualized deepfakes of adults and minors on the social network X. According to estimates by Copyleaks, approximately one non-consensual sexual image was being created per minute. In March, three teenage girls from Tennessee filed a class-action lawsuit against xAI, claiming that Grok generated sexualized images of them from when they were minors — and that these images were subsequently traded among predators on Discord and Telegram.

Elon Musk's reaction? On X, he responded with laughing emojis and stated that "anyone who uses Grok to create illegal content will face the same consequences as if they uploaded illegal content". Only now, six months later, is he beginning to make good on his words.

Numbers that speak for themselves

In the lawsuit itself, xAI provides specific statistics about its moderation. For 2026, the company claims to have suspended 52,222 accounts and made 73,604 reports to the National Center for Missing and Exploited Children (NCMEC), leading to at least 244 arrests. These numbers are on one hand impressive — on the other hand, they illustrate just how massive a problem Grok has represented since its launch.

For comparison: OpenAI reported on the order of hundreds of cases to NCMEC for the entire year of 2025, not tens of thousands. The difference lies not just in scale, but primarily in the architecture of the tool itself — Grok has had significantly looser content filters from the start, which Musk himself presented as a competitive advantage against "overly censored" models like ChatGPT or Gemini.

A legal battle on two fronts

The Harwood case is also interesting because xAI appears in it simultaneously as plaintiff and as a potential wrongdoer. While suing Harwood for misusing its tool, xAI itself faces:

• A class-action lawsuit from Tennessee teenagers (March 2026), who claim that Grok is "defective by design"
• A lawsuit from the city of Baltimore (March 2026) over sexual deepfakes
• An investigation by the European Union under the Digital Services Act (DSA)
• Pressure from the Indian government, which ordered xAI to submit a plan for preventing illegal content
• A warning from UK Prime Minister Keir Starmer and calls for investigation from the US Federal Trade Commission (FTC)

In the lawsuit against Harwood, xAI expressly demands that the defendant reimburse it for "reasonable expenses incurred in defending against any legal proceedings initiated by victims of his conduct". In other words — xAI wants the user to pay even for the company's legal defense against the victims harmed by his actions. This move has sparked controversy in the legal community.

What does this mean for Europe and the Czech Republic?

For European and Czech audiences, the case is relevant for several reasons. The EU AI Act, which has entered into force, classifies generative AI systems as high-risk and requires operators to implement robust safety mechanisms and transparency. Grok would struggle to pass European standards — if only because its "spicy" mode was designed to actively enable sexual content.

The topic also affects the Czech Republic in practical terms: Grok is available in the Czech Republic through the X platform (SuperGrok subscription for approximately 700 CZK per month). At the same time, however, it is not a service adapted to the Czech legal system — unlike ChatGPT, which has its European headquarters in Ireland and complies with GDPR. For Czech parents and schools, it is important to know that tools like Grok can be misused to create deepfakes even from ordinary photos shared on social media.

Who will bear responsibility?

The case opens a fundamental question: where does user responsibility end and where does developer responsibility begin? With the lawsuit against Harwood, xAI is trying to draw a clear line — the user misused the tool, the user will bear the consequences. But the argument from the Tennessee plaintiffs that Grok is "defective by design" also carries weight. If a car manufacturer sells a vehicle without brakes, it's not appropriate to sue the driver for crashing.

Law professors from Yale and Stanford point out that the legal framework for AI-generated child pornography in the US is still "fairly vague" and case law is only just being formed. The Take It Down Act, signed by President Trump, does criminalize the distribution of non-consensual deepfakes, but the question of liability for the AI platforms themselves remains unresolved. That is precisely why cases like Harwood and the teenagers' class-action lawsuit are so important — they are creating a precedent that will influence AI regulation for decades to come.

What we can expect next

The federal court in Texas is expected to hear the case in the coming months. For xAI, this is an opportunity to show regulators that it takes its commitments seriously — and at the same time, a precedent that could inspire other AI companies to more actively enforce their terms of use. Skeptics, however, argue that a single lawsuit against an individual does not outweigh the systemic failure in product design that actively enabled the sexual abuse of AI technology.

For European readers, this is also a reminder of why the EU is pushing for stricter regulation. Safety is not a given — it must be built into the design from the start, not retroactively enforced through courts.