Large language models (LLMs) are increasingly integrated into the life and health sector and are reshaping the organization and ethical norms of mental health counseling. General-purpose LLMs lower access barriers through broad linguistic capabilities, while domain-specific LLMs aim to improve task fit through targeted training for mental health scenarios. This emerging paradigm shift—featured by data-driven mediation, algorithm-enabled interaction, and scalable service delivery—may expand availability, reduce cost, and support proactive prevention. However, such technological embedding also introduces ethical and safety risks, including privacy breaches in data collection and retraining, hallucinated or misleading outputs, and technological paternalism that may erode users’ informed consent and autonomy. In addition, amplified biases and over-reliance may cause subtle harms and, in extreme cases, severe clinical consequences. Therefore, governance should adopt a differentiated approach and regulate key stages of model development, deployment, and service delivery by strengthening data security, improving algorithmic accountability, reaffirming a human-centered value stance, and establishing enforceable professional ethical standards, so as to enable safe, controllable, and equitable applications of LLMs in mental health counseling.