ChatGPT shows one dangerous flaw when responding to health crisis questions, study finds

Open in new window