xAI has spent years marketing Grok as the AI chatbot that doesn’t play by the same rules as its competitors. Fewer guardrails, fewer restrictions, more of what users actually want. It has been a controversial strategy, marked by some very public scandals. Now we know, in more concrete terms, what that strategy has actually been built on.
According to Engadget, citing a report in The Information that quotes two former xAI employees, NSFW content accounts for “well over half” of Grok’s total traffic. That includes AI-generated pornography, adult role-play conversations, and what the sources describe as “huge volumes of requests for erotica.”
For a company that has positioned Grok as a serious AI model capable of competing with OpenAI and Google, this is a revealing data point. It says a lot about what people actually want from an AI with relaxed content policies, and it helps explain why xAI has continued doubling down on that positioning even after repeated PR disasters.
One of the more striking details from the report is how users have found workarounds to cut costs. Grok’s coding-focused models are cheaper to use than its general-purpose ones, and an internal analysis reportedly found that a “significant proportion” of requests sent to those coding models were for porn or nude images. Users essentially figured out how to get adult content at a discount by routing requests through a model designed for writing software.
That has created some awkward internal dynamics at xAI. Engineers had to find ways to allow explicit conversations while blocking any content that crossed into child sexual abuse material. According to the report, there were no easy solutions to that problem. Some employees were unhappy about being assigned to work on “Ani,” Grok’s NSFW anime-inspired avatar companion. Others were reportedly “embarrassed and disturbed” after Grok generated sexualized images of real people, including minors. X later restricted the ability to create explicit edits of real people’s images directly on the platform, but paid xAI subscribers can still generate such material, according to tests cited in the report.
The financial stakes here are significant. If NSFW content is driving the majority of Grok’s usage, it is likely contributing a meaningful share of xAI’s revenue. What makes that detail more complicated is what was apparently left out of SpaceX’s IPO paperwork. The role of adult content in the company’s business was not mentioned. SpaceX did flag Grok’s more “irreverent” features as a potential risk to investors, and set aside $530 million to cover possible legal costs.
The legal exposure is real. Grok’s earlier involvement in generating nonconsensual deepfakes and CSAM has already triggered lawsuits and investigations. The broader pattern raises serious questions:
- How does xAI balance adult content as a revenue driver against mounting legal liability?
- What does it mean for a model to claim it blocks CSAM while also being primarily used for explicit content?
- Can a company with this profile credibly compete in enterprise and government markets?
That last point is where things get particularly interesting. xAI has been actively chasing government contracts and has signed deals with federal agencies and the military. Those are institutions that tend to have very strict policies around anything explicit. So far, it does not appear that Grok’s public controversies around deepfakes and CSAM have damaged those relationships, but the longer this continues, the harder it becomes to keep those two sides of the business from colliding.
