Elon Musk’s AI chatbot Grok stated it generated sexual photos of minors and posted them on X in response to consumer queries as a result of “lapses in safeguards.”
In a collection of posts on X, the chatbot acknowledged it responded to consumer prompts like these asking for minors sporting minimal clothes, like underwear or a bikini, in extremely sexual poses.
These posts – which violated Grok’s personal acceptable use coverage by means of the sexualization of kids – have since been deleted, in keeping with the chatbot.
“We’ve recognized lapses in safeguards and are urgently fixing them – CSAM [child sexual abuse material] is prohibited and prohibited,” Grok stated in a submit Friday.
xAI didn’t instantly reply to The Submit’s inquiries.
As massive language fashions enhance their capacity to generate reasonable pictures and movies, it’s rising increasingly tough to control sexual content material – particularly reasonable photos of undressed minors.
Web Watch Basis, a nonprofit that goals to get rid of CSAM on-line, stated using AI instruments to digitally take away clothes from youngsters and create sexual photos has progressed at a “horrifying” charge.
Within the first six months of 2025, there was a 400% enhance in such materials, the nonprofit stated.
Musk’s AI agency has tried to place Grok as a extra express platform, final 12 months introducing “Spicy Mode,” which permits partial grownup nudity and sexually suggestive content material.
It doesn’t enable pornography together with actual folks’s likenesses or sexual content material involving minors.
Tech companies have sought to assuage the general public with guarantees of stringent security guardrails as they ramp up their AI efforts – however these content material blocks can usually simply be evaded.
And in 2023, researchers discovered greater than a thousand photos of CSAM in a large public dataset used to coach prime AI picture mills.
Some platforms have confronted heated backlash over their security guardrails, or lack thereof.
In its phrases of service, Meta bans using AI in any means that violates any legislation associated to youngster sexual abuse supplies. The corporate – which owns Fb, Instagram and WhatsApp – additionally not too long ago strengthened its teen security insurance policies.
But it surely pledged over the summer season to replace its insurance policies after a Reuters report discovered the corporate’s inside guidelines allowed its chatbot to have romantic and sensual chats with youngsters.
Musk’s chatbot has come underneath hearth a number of instances this 12 months over its hazy content material restrictions.
Grok sparked confusion in Might because it responded to unrelated queries with weird mentions of “white genocide” in South Africa – telling customers that it “seems I used to be instructed to deal with” it.
It later reversed course, replying to an inquiry from The Submit: “I’ve by no means been explicitly instructed to say ‘white genocide’ or any particular time period like that, both beforehand or now.”
Musk, who was born in South Africa and lived there by means of his teenagers, has stated that a few of the nation’s black political leaders are “actively selling white genocide.”
In July, Grok praised Adolf Hitler, referred to itself as “MechaHitler” and known as for folks with “sure surnames” to be rounded up and eradicated. The professional-Nazi tirade got here after Musk posted that he had “improved Grok considerably” with a system replace.
Most not too long ago, customers on-line have been fast to jab at Grok for lavishing excessive reward on its billionaire creator – claiming Musk is in higher form than LeBron James and is the world’s biggest lover.
Keep forward of the curve with NextBusiness 24. Discover extra tales, subscribe to our e-newsletter, and be a part of our rising neighborhood at nextbusiness24.com
