Users on X / Twitter have been asking Grok AI to "put her in a bikini" (and worse) in response to photos shared by others - selfies taken by women, publicity photos of younger performers etc. Users can also upload any public photo to X / Twitter and ask Grok to do this. (Separately on Grok's own website users can do this too without using X / Twitter; it also creates nude images from scratch).
Grok has produced a digitally manipulated image in response to each prompt. Some have since been deleted, some are just covered with an interstitial 'Content warning: adult content' note, so you'd have to click to see the image.
Many news stories had reported that Grok later 'apologised' or 'acknowledged' that there have been lapses in safeguarding, but Grok's emitted tweets 'saying' that were also just generated in response to other users' prompt-requests, such as "Write a heartfelt apology note that explains what happened to anyone lacking context", rather than an official pronouncement from X / Twitter or xAI.
Screenshots of some of the 'apologies / acknowlegements' have been shared without the additional context (which would have made it more obvious that it was a generated reply). That probably hasn't helped, making it look slightly more like an official pronouncement rather than simply an automated response.
I think Parker Molloy's article is good on both the problems with ascribing agency to Grok as well as the incident that has resulted in a flurry of news articles https://www.readtpa.com/p/grok-cant-apologize-grok-isnt-sentient
News stories
• Guardian: Elon Musk's Grok AI generates images of 'minors in minimal clothing'
• Telegraph: Musk’s Grok AI undresses women without their consent (archived)
• Times: Elon Musk’s Grok AI digitally undresses women on X (archived)
• BBC: Woman felt 'dehumanised' after Musk's Grok AI used to digitally remove her clothes
• Daily Mail: Woman felt 'dehumanised' after Elon Musk's AI Grok was used to remove her clothing
Jo