According to Sentinel Beating monitoring, several former employees revealed that Elon Musk's AI company xAI is increasing its investment in Grok's video and image generation capabilities. Two recently departed employees estimated that over half of Grok's traffic comes from adult images, videos, and role-playing dialogues, among other NSFW activities.
Even Grok's code generation model is extensively used to produce adult content, mainly due to the lower cost of running the code model. NSFW subscriptions have brought significant cash flow to xAI, with the premium subscription package Grok Imagine charging as high as $300 per month. SpaceX's IPO prospectus disclosed that its AI division is expected to generate $3.2 billion in revenue in 2025, scaling up to generating 10 billion images and 2 billion videos per month in the first quarter of 2026.
However, the relaxation of restrictions has also sparked regulatory and internal disagreements. The EU, France, and the UK have launched investigations into Grok's generation of underage pornographic images. Within xAI, the generation of real human nude images has raised concerns among researchers, leading to the departure of some technical staff who were compelled to work on developing the sexualized avatar "Ani." SpaceX cautioned in its prospectus that Grok's capacity to generate explicit and unrestrained content could pose risks to reputation damage and the exposure of graphic images.
In the long run, Grok's video labeling data has also been involved in training the "world model" for predicting real-world scenarios, potentially for use in Tesla's self-driving technology and Optimus robots. In contrast, competitors have strictly controlled adult content. Anthropic explicitly prohibits Claude from generating sexually explicit content, while Google and Midjourney's video and image models similarly block adult content. OpenAI has also postponed its plans for erotic content.
