Functionally, it probably amounts to two restrictions: a chatbot cannot formally diagnose & a chatbot cannot bill insurance companies for services rendered.

Most "therapy" services are not providing a diagnosis. Diagnosis comes from an evaluation before therapy starts, or sometimes not at all. (You can pay to talk to someone without a diagnosis.)

The prohibition is mainly on accepting any payment for advertised therapy service, if not following the rules of therapy (licensure, AI guidelines).

Likewise for medicine and law.

After a bit of consideration I’m actually ok with codifying Bad Ideas. We could expand this.