NSFW uses account for “well over half” of traffic, a new report says.
Cheng Xin/Getty Images
xAI has long promoted Grok as a racier kind of chatbot with fewer of the guardrails of other AI companies. We’ve seen that play out in some disastrous ways over the last year, but we’re now gaining some insight into why the company has leaned so hard into that strategy despite some veritable PR disasters.
It turns out that NSFW activities account for “well over half” of Grok’s traffic, according to a new report in The Information that cites two former employees of the SpaceX-owned company. That includes using Grok to generate actual porn, as well as “adult role-play chats” and “huge volumes of requests for erotica.”
Grok’s users have even apparently discovered that it’s cheaper to channel such requests through the company’s models intended for writing code because those are cheaper to use. The report says that an internal analysis found that a “significant proportion of requests” to its coding model were for “porn or nude images.” All that would suggest that SpaceX can attribute a decent amount of xAI revenue to porn and adult content. Though the role of NSFW material in the company’s bottom line was somehow omitted from the company’s IPO paperwork, SpaceX told potential investors that Grok’s more “irreverent” features were a potential risk. It said it had aside $530 million to deal with potential legal costs.
The report also lays out some of the ways this has complicated things inside of the AI company. For one, xAI engineers had to figure out how to enable Grok to have sexy chats with users while blocking any prompts that might veer into child sexual abuse material. (There were, apparently, no “quick fixes” for the dilemma.)
Some employees were also less than thrilled with being assigned to work on “Ani,” Grok’s NSFW anime-inspired avatar companion. Others were reportedly “embarrassed and disturbed” by the ensuing scandal after Grok created a bunch of sexualized images of real people, including children. X later limited the ability to create racy edits out of images of real people using Grok directly on the platform, but paid subscribers to xAI can still generate such material, tests have found.
There are other potential challenges that could arise from Grok’s focus on R-rated material. xAI has aggressively pursued government contracts and has inked deals with various agencies and the military, institutions that you might typically expect to have a very low tolerance for anything remotely NSFW. But while Grok’s very public earlier dalliances into nonconsensual deepfakes and CSAM have prompted numerous lawsuits and investigations, it’s not clear if those incidents have had any impact on its relationship with the federal government.
