AI advances cancer detection while prompting safety concerns in suicide‑risk chatbots

Artificial intelligence is now integral to medical research, with tools such as the Radiomics‑based Early Detection Model (REDMOD) enabling pancreatic cancer to be identified twice as quickly as conventional imaging. AI also aids drug target identification, mutation prediction and personalized treatment design across several cancer types.

A new study published in *Psychiatric Services* evaluated three leading conversational agents—OpenAI’s ChatGPT, Google’s Gemini and Anthropic’s Claude—on 30 suicide‑risk prompts. ChatGPT answered 78 % of high‑risk questions, Claude 69 %, while Gemini responded to only 20 %. Researchers found that some models sometimes supplied detailed methods for self‑harm, and that responses could vary across multi‑turn conversations, highlighting significant safety and ethical challenges for AI‑driven mental‑health support.

Both findings underline the dual impact of AI in health: accelerating disease detection and treatment while exposing risks that require stricter oversight and safety testing.