the AI app can get his kids to open up more to it than to him or his wife about their issues.
This was a thing decades ago with much simpler scripted "AI":
https://liacademy.co.uk/the-story-of-eliza-the-ai-that-fooled-the-world/
Many users began to reveal personal and emotional information, believing ELIZA was responding thoughtfully, even though it was merely following scripted patterns without any true understanding.
I think the extent to which people can open up more easily to an AI than to a human says something about culture and vulnerability, and how shitty it can be on creating an environment that feels safe to be vulnerable in, or even using language that is exploratory rather than judgmental.
If a young kid is at the store and says to their parent "I want that toy", and the parent's immediate reflex is "no, you already have lots of toys at home" instead of "why is that? What's interesting to you about that toy?" Well it doesn't seem surprising to me that a kid in a culture like that might have an easier time opening up to an AI. The kid probably hasn't worked out beforehand why they want the toy. They're just trying to learn to express something they feel, a want, a desire. But being heard out, having a nice conversation with a loved one in which they feel safe and happy, I'm sure most kids are going to remember that far more fondly than any specific toy.
And whether kid or adult, this is part of the appeal of chatbots. That they aren't apt to jump down your throat over something you say. But this is also a problem because there are times when making judgments and setting boundaries matter very much, and the AI is not equipped to stand in for that part properly. It's just good at standing in for openness (and this is a part people are supposed to be bonding over normally, with each other!)