In a recent Forbes analysis, AI expert Dr. Lance B. Eliot examines a proposal aimed at reshaping how artificial intelligence systems handle mental health support. Right now, large language models (LLMs) like ChatGPT, Claude, Gemini, and others often end up giving mental health guidance — even though that isn’t their primary design purpose and they lack clinical training. This has led many users to turn to general-purpose AI for emotional support and life advice, blurring the lines between casual interaction and professional mental health care.
The idea being floated would require major AI developers to remove all built-in mental health advice capabilities from their general models and instead integrate specialized therapist-oriented LLMs designed specifically for that purpose. Proponents argue this could reduce legal and safety risks by ensuring users seeking mental health assistance are directed to systems better equipped — at least in theory — to provide such guidance, rather than relying on generic AI that may give inaccurate or unhelpful responses.
However, implementing such a carve-out poses technical challenges. Generic AI models treat all conversational topics as interconnected, so removing any mental health-related knowledge isn’t as simple as flipping a switch. If the general model suddenly lacks context about a user’s emotional state, a separate mental-health LLM might start from scratch and give inconsistent or even contradictory advice. That could leave users confused or mistrustful, undermining the intended safety benefits.
Critics also note broader questions about whether AI should be involved in mental health support at all, given that many states already pass laws restricting AI from providing diagnosis or therapeutic decisions. For example, jurisdictions like Nevada and Illinois have banned AI tools from making clinical mental health decisions, highlighting ongoing regulatory uncertainty about how best to integrate or limit such technologies.
