Elon Musk’s AI chatbot Grok said it generated sexual images of minors and posted them on X in response to user queries due to “lapses in safeguards.”
In a series of posts on X, the chatbot acknowledged it responded to user prompts like those asking for minors wearing minimal clothing, like underwear or a bikini, in highly sexual poses.
Those posts – which violated Grok’s own acceptable use policy through the sexualization of children – have since been deleted, according to the chatbot.
“We’ve identified lapses in safeguards and are urgently fixing them – CSAM [child sexual abuse material] is illegal and prohibited,” Grok said in a post Friday.
xAI did not immediately respond to The Post’s inquiries.
As large language models improve their ability to generate realistic photos and videos, it is growing more and more difficult to regulate sexual content – specifically realistic images of undressed minors.
Internet Watch Foundation, a nonprofit that aims to eliminate CSAM online, said the use of AI tools to digitally remove clothing from children and create sexual images has progressed at a “frightening” rate.
In the first six months of 2025, there has been a 400% increase in such material, the nonprofit said.
Musk’s AI firm has tried to position Grok as a more explicit platform, last year introducing “Spicy Mode,” which allows partial adult nudity and sexually suggestive content.
It does not allow pornography including real people’s likenesses or sexual content involving minors.
Tech firms have sought to assuage the public with promises of stringent safety guardrails as they ramp up their AI efforts – but these content blocks can often easily be evaded.
And in 2023, researchers found more than a thousand images of CSAM in a massive public dataset used to train top AI image generators.
Some platforms have faced heated backlash over their safety guardrails, or lack thereof.
In its terms of service, Meta bans the use of AI in any way that violates any law related to child sexual abuse materials. The company – which owns Facebook, Instagram and WhatsApp – also recently strengthened its teen safety policies.
But it pledged over the summer to update its policies after a Reuters report found the company’s internal rules allowed its chatbot to have romantic and sensual chats with children.
Musk’s chatbot has come under fire several times this year over its hazy content restrictions.
Grok sparked confusion in May as it responded to unrelated queries with bizarre mentions of “white genocide” in South Africa – telling users that it “appears I was instructed to address” it.
It later reversed course, replying to an inquiry from The Post: “I’ve never been explicitly instructed to mention ‘white genocide’ or any specific term like that, either previously or now.”
Musk, who was born in South Africa and lived there through his teens, has said that some of the country’s black political leaders are “actively promoting white genocide.”
In July, Grok praised Adolf Hitler, referred to itself as “MechaHitler” and called for people with “certain surnames” to be rounded up and eliminated. The pro-Nazi tirade came after Musk posted that he had “improved Grok significantly” with a system update.
Most recently, users online were quick to jab at Grok for lavishing extreme praise on its billionaire creator – claiming Musk is in better shape than LeBron James and is the world’s greatest lover.
Credit: Source link
