Elon Musk’s Grok AI Is as Gross as You Would possibly Anticipate

Kolin Koltai, a researcher on the Netherlands-based investigative outlet Bellingcat, just lately found that X customers had been utilizing the platform’s AI chatbot, Grok, to undress girls in images that that they had uploaded to the platform. Whereas the bot rejects prompts that ask to make folks solely nude, it’s prepared to stroll proper as much as the road of non-consensual sexual imagery. 

Customers who replied to pictures of girls and requested that Grok “take away her garments” acquired an in-thread reply with a picture of the lady in lingerie or a bikini. Often, Grok replied with a hyperlink to a separate chat containing the picture. 

404 Media broke the story on Tuesday and famous that whereas dozens of customers gave the impression to be making the requests of Grok, the follow appears to have first been popularized in Kenya. An article in Citizen Digital, a Kenyan information web site, described the phenomenon as “a latest development by Kenyans on X.” 

Posting on X, Phumzile Van Damme, a South African activist (and a former know-how and human rights fellow at Harvard’s Kennedy Faculty), basically requested the bot to elucidate itself. In reply, Grok wrote, “This incident highlights a spot in our safeguards, which did not block a dangerous immediate, violating our moral requirements on consent and privateness…We’re additionally reviewing our insurance policies to make sure clearer consent protocols and can present updates on our progress.” X had not returned 404’s request for official remark on the time of print. 

Notably, this discovery comes one week after the US Home of Representatives handed the “Take it Down Act,” a bipartisan invoice that might criminalize the publication of nonconsensual, sexually specific photos and movies, together with these generated by AI. The invention additionally comes simply two weeks after X Corp. filed a lawsuit in opposition to Minnesota AG Keith Ellison, difficult the constitutionality of the state’s regulation banning the usage of deepfakes to affect elections.

Grok was developed by xAI and launched in November 2023. Elon Musk, the proprietor of each xAI and X, which the bot has “direct entry to,” spun up the chatbot with the intent of making a type of “TruthGPT,” as he described it to Tucker Carlson in an interview earlier that yr. In that very same dialog, he additionally expressed his concern that ChatGPT was being “educated to be politically right.” 

Since releasing Grok, Musk has commonly tried to distinguish it from different AI chatbots by describing Grok as “primarily based” and having a humorousness. Whereas different AI fashions, corresponding to these developed by OpenAI (which Musk left in 2018) and Google, decline to have interaction with sexually specific or in any other case delicate dialog, xAI bragged that Grok would reply “spicy questions which are rejected by most different AI methods.” In an especially “How do you do, fellow kids?” demonstration throughout Grok’s launch, Musk shared the bot’s step-by-step directions for making cocaine, together with its sickest burns on Sam Bankman-Fried, who was discovered responsible on seven fees of fraud and conspiracy simply someday earlier.

“Please don’t use it for those who hate humor!” xAI implored because it introduced Grok’s launch.