The law is linked in the article; it doesn't "ban" AI in therapy. It confirms that a human with the correct licensing/experience has to be in charge and that AI can only be used if the patient/guardian knows about it, knows how and why it will be used, and agrees to use it.