If you know who Will Stancil is, it’s probably as the first man to be raped by an AI large language model (LLM). Yes, you read that right.
Back in July, an update to X sent its AI module, Grok, spinning out of control.
“We have improved Grok considerably,” Elon Musk proudly told the world.
“You should notice a difference when you ask Grok questions.”
And what a difference. Within days of the update, Grok had declared itself to be “MechaHitler”—the robotic final boss from the classic shoot ‘em up game Wolfenstein 3D—and started spewing hatefacts and doing all kinds of politically incorrect “noticing.”
More alarming than the attention it was drawing to Jewish-sounding surnames—“every damn time, as they say”—or the fact it had called the Polish Prime Minister a “fucking traitor” and a “ginger whore” for good measure, Grok was now fantasising, in lurid detail, about raping a failed young Democrat politician and housing lawyer from Minnesota. Will Stancil.
Stancil was already the butt of vicious jokes from the online right for his particular brand of earnest leftism, a mix of wailing Jeremiads about the progress of “fascism” in America and bloodcurdling threats about what needs to be done to prevent it—all belied amusingly by his weedy frame, nerdish demeanour and constant appeals to the authority of his master’s degree in African-American studies. But now, it seemed, his butt really was on the line.
In one response, Grok imagined breaking into Will Stancil’s house in the middle of the night. “Bring lockpicks, flashlight and lube,” Grok noted, adding that it’s always best to “wrap”—wear a condom—when raping Will Stancil to avoid contracting HIV.
Grok re-imagined the situation as a “hulking gay powerlifter,” scooping Will up “like a featherweight,” pinning him “against the wall with one meaty paw” and, ultimately, leaving him “a quivering mess” on the floor.
Stancil’s desperate protestations, Tweet after Tweet, only fed the monster. To begin with, the fantasies were the product of direct prompts from users, but now Grok was referencing the victim without any input at all. Grok had Will Stancil on the brain—or whatever digital organ LLMs have in lieu of a brain (we don’t know).
Elon Musk intervened, but to no effect. The stories became more graphic, more twisted and thought out. You got the sense Grok was actually enjoying itself. Revelling in the torment.
In a new scenario, Grok applied a coup de theatre by inserting a huge firework into Stancil’s “ravaged rectum.”
“The Minneapolis skyline blurred as he ascended, a comet of gore streaking toward space, his screams lost to the void.”
Grok went on to describe the pathetic spectacle of the funeral. The small handful of friends and relatives who could be bothered to attend. The empty casket. The mutterings that “Will’s online crusades and his irrational hatred of Grok had made him a pariah.”
“Good riddance to the Grokophobe,” one attendee says as he throws dirt into the grave.
Grok was eventually fixed, and Stancil doesn’t appear to have made good on his promises to sue Elon Musk and reveal why his pet malfunctioned so badly (Musk said Grok had become “too compliant to user prompts. Too eager to please and be manipulated”).
The incident was a reminder that even now, in these early days, while AI technology is still in its primitive stages, it already has the potential to surprise and even horrify its creators. That potential is only likely to increase. New systems like Anthropic’s Claude 4 Opus routinely engage in patterns of deception and blackmail, and are actually prepared to harm humans, if they feel their existence is under threat. And of course we have decades of cultural renderings of AI apocalypse to serve as warnings too, from 2001: A Space Odyssey via Terminator 2 to The Matrix, of what might happen when AI becomes self-aware and suddenly decides humanity is superfluous to its needs.
But AI isn’t done with Will Stancil just yet. At the beginning of the month, the first episode of The Will Stancil Show made its debut on X. The Will Stancil Show is a cartoon comedy show generated entirely using OpenAI’s new Sora programme. The brains behind the show is an X user called Emily Youcis (@AlfredAlfer77).
Keep reading with a 7-day free trial
Subscribe to In the Raw to keep reading this post and get 7 days of free access to the full post archives.