I had to make a choice to not even use Grok (I wasn't overly interested in the first place, but wanted to review how it might compare to the other tools), because even just the Explore option shows photos and videos of CSAM, CSAM-adjacent, and other "problematic" things in a photorealistic manner (such as implied bestiality).

Looking at the prompts below some of those image shows that even now, there's almost zero effort at Grok to filter prompts that are blatantly looking to create problematic material. People aren't being sneaky and smart and wordsmithing subtle cues to try to bypass content filtering, they're often saying "create this" bluntly and directly, and Grok is happily obliging.