“We are aware of recent posts made by Grok and are actively working to remove the inappropriate posts,” Grok’s maker xAI said on the X social media platform.

But the episode was more than that: It was a glimpse of deeper, systemic problems with large language models, or L.L.M.s, as well as the enormous challenge of understanding what these devices really a...