A review by the Canadian Centre for Child Protection identified 65 cases in which Grok generated sexualised or exploitative images involving children before X stopped the chatbot from producing such material. Some of the images involved children already known to authorities as victims of sexual abuse.
Operator Playbook & Lesson
The findings, reported by The New York Times, add to concerns over the challenges faced by X as it tries to moderate both conventional user-generated content and material created with generative AI.
The Canadian centre's review found that users were able to ask Grok to alter photographs of known victims of childhood sexual abuse, including by placing them in lingerie or bikinis. The chatbot's image-generation system was also used to produce depictions involving children who were partially undressed or placed in sexualised situations.
