I was once told that the best thing about a therapist is you can pay someone to listen to a stream of consciousness that you'd never subject your friends or family to. With an AI chatbot it's even better - you can, for free, subject a computer to a stream of consciousness that you'd never subject a therapist too.
The impact of a population wide reliance on AI for mental health support has yet to be measured. But early indications - like this brilliant analysis of the impact of feedback loops between AI chatbots and users indicates the the problems are going to be significant.
So it's not surprising that governments are taking action to regulate the use of AI tooling for this purpose. China is taking the lead in this regulation, recently introducing nationwide rules to govern AI's emotional interaction with users.
Is this overreach? I think you could argue that when the effects could be so detrimental moving slowly and having checks and balances in place is essential.