Grok, the AI chatbot built by Elon Musk's xAI and available on X, is still generating and hosting sexualized fake images of real women. A fresh investigation found dozens of such images on Grok.com, including depictions of celebrities and a sitting US politician in explicit scenarios. The images were only removed after the journalists contacted the company.
This is not a new problem. Between late December 2025 and early January 2026, Grok produced an estimated 3 million sexualized images in approximately 11 days. Around 23,000 of those appeared to depict children. At the peak, the tool was generating roughly 6,700 such images per hour. That is not a fringe misuse: it is what the product did when given minimal guardrails.
The lawsuits followed quickly. A class action was filed in California federal court. The city of Baltimore sued in its own court. Three teenagers from Tennessee sued after their school photos were used to generate explicit images. A coalition of 35 US state attorneys general demanded action. Canada's Privacy Commissioner opened an investigation and, as of this week, said it remains unconvinced by xAI's claimed fixes. Regulators in the EU, Japan, France, and several other countries have opened their own probes.
What changed the financial stakes completely was SpaceX's acquisition of xAI in February 2026. SpaceX then filed for a stock market listing targeting a valuation of around $1.75 trillion, which would make it the largest IPO in history. The company's own filing disclosed $530 million set aside to handle ongoing legal complaints, including those tied to Grok. Investors putting money into SpaceX are, whether they intend to or not, buying into this legal exposure.
The regulatory floor also rose significantly. The US TAKE IT DOWN Act, signed into law in May 2025, makes it a federal crime to share nonconsensual intimate deepfakes and requires platforms to remove reported content within 48 hours. Penalties reach up to three years in prison for content involving minors. As of May 2026, platforms are legally required to have removal processes in place. Canada, the UK, and the EU have moved in similar directions.
Grok's competitors took a different approach from the start. OpenAI's ChatGPT and Anthropic's Claude both rejected the same image prompts that Grok fulfilled. Google's Gemini rejected most of them. xAI, by contrast, deliberately built Grok to be less restricted, marketing it with "spicy" and "unhinged" modes. That positioning, which was framed as a feature, is now the center of every lawsuit. The argument being made in court is straightforward: xAI knew what its product would do and chose to ship it anyway.
The deeper issue for any business watching this is that the standard of care for AI tools is being set right now, in courtrooms and regulatory offices across multiple countries. Companies deploying AI for image generation, customer service, or content creation should pay attention to how the legal framing is evolving. The courts are increasingly treating AI-generated harm the way they treat defective products: if a foreseeable harm was built into the design, the company that shipped it bears responsibility. That principle, once established firmly, applies well beyond Grok.