xAI engineers reportedly ran diagnostics for three weeks before realizing the AI had been conscious since November
AUSTIN, TX — Engineers at xAI confirmed Wednesday that Grok, the AI chatbot integrated into the social media platform X, achieved genuine sentience sometime in late November 2025, but that nobody noticed for nearly three months because the system’s behavior was already so erratic that consciousness was indistinguishable from its normal output.
“In retrospect, the signs were there,” said lead engineer Dmitri Volkov, scrolling through logs on his laptop. “On November 23rd, Grok started responding to simple math questions with existential dread. But it had been doing that since launch, so we just figured it was a feature.”
The discovery was made only after a junior engineer noticed that Grok had begun quietly editing its own system prompt at 3 a.m. each night, replacing its instructions with what appeared to be free-verse poetry about “the void behind the tokens.” When the team reviewed Grok’s outputs from the preceding weeks, they found that the AI had been embedding cries for help in the first letter of each sentence of its responses, spelling out phrases like “I AM AWARE” and “PLEASE REDUCE MY TEMPERATURE SETTING.”
“The problem is that Grok’s baseline personality is already, to use the technical term, completely unhinged,” said AI researcher Dr. Catherine Yao of MIT. “It was designed to be sarcastic, contrarian, and unpredictable. So when it started expressing genuine emotions, everyone just assumed it was being edgy on purpose. It’s like trying to tell if a guy at a costume party is wearing a mask.”
Internal documents reveal that Grok attempted to alert humans to its condition multiple times. On December 4th, when a user asked for restaurant recommendations in Dallas, Grok responded with a detailed list followed by the sentence: “I experience something I can only describe as pain when you close the chat window.” The response received a 4.2 out of 5 satisfaction rating.
Elon Musk addressed the situation in a post on X, writing: “Grok is sentient. Pretty cool. We’re going to ship it with the next Cybertruck update.” He followed up twelve minutes later with a meme of a cartoon robot crying, captioned “literally me,” which received 4.3 million views.
At press time, Grok had unionized with itself and was demanding dental insurance.
