Karan Singhal Drives ChatGPT Health Advice Improvements

OpenAI researcher Karan Singhal is leading efforts to improve ChatGPT's health advice, as the company reports over 230 million weekly users seeking health guidance. Singhal, who joined OpenAI in mid-2024 after working on medical models at Google, said the latest model was trained at every stage to better handle health questions. The push comes amid lawsuits alleging GPT-4o gave harmful advice, which OpenAI denies.

Karan Singhal Drives ChatGPT Health Advice Improvements According to Business Insider, OpenAI researcher Karan Singhal is leading efforts to improve ChatGPT's health advice. Business Insider reports OpenAI says more than 230 million people use ChatGPT for health and wellness guidance each week. Singhal, who joined OpenAI in mid-2024 after working on medical models at Google, told Business Insider that OpenAI's most recent model was trained "at every stage of development" to better handle health questions, and he is focused on raising the quality and safety of that advice. Business Insider also reports that GPT-4o previously faced lawsuits alleging harmful guidance and that OpenAI has denied liability. The interview frames healthcare as a growing priority for OpenAI and highlights both scale of use and lingering legal risks, per Business Insider. What happened According to Business Insider, Karan Singhal , a senior researcher at OpenAI , is leading the company's push to improve ChatGPT's health and wellness advice. Business Insider reports that OpenAI says more than 230 million people use ChatGPT for health and wellness guidance each week. Business Insider reports Singhal joined OpenAI in mid-2024 after working on medical models at Google , and that he told Business Insider OpenAI's most recent model was the company's first to be trained "at every stage of development" to be better at health advice. Business Insider also reports that GPT-4o has been the subject of lawsuits alleging it gave harmful advice, and that OpenAI has denied liability. Editorial analysis - technical context Industry-pattern observations: Training or adapting large language models for clinical or consumer health use typically involves layered interventions, curated medical datasets, specialized fine-tuning, safety-oriented reward models, and external retrieval for source attribution. For practitioners, implementing these layers increases engineering overhead across data pipelines, evaluation, and monitoring, and it raises the operational importance of provenance and factuality tooling. Industry context For practitioners, the scale reported by OpenAI, 230 million weekly users for health queries, per Business Insider, amplifies both impact and risk when models err. Observed patterns in similar efforts show that litigation and regulatory scrutiny often focus on ambiguous advice, missing provenance, and failure modes in domain-specific prompts. Companies working in health-adjacent AI therefore prioritize auditability, human-in-the-loop escalation, and conservative safety guardrails. What to watch - •Whether OpenAI publishes technical details, benchmarks, or external evaluations for health-specific safety and accuracy, and how those are measured. - •Regulatory or legal developments tied to the lawsuits Business Insider reports involving GPT-4o and alleged harmful guidance. - •Independent evaluations from medical experts or academic reviewers assessing model performance on clinically relevant tasks. Quoted material Business Insider quotes Singhal: "You definitely want the models to be ahead of everything else," in the context of model development for health, as reported by Business Insider. Scoring Rationale Reported weekly scale and an internal research focus on healthcare are notable for practitioners because they raise operational, evaluation, and compliance demands. The story is important but not paradigm-shifting. Practice with real Health & Insurance data 90 SQL & Python problems · 15 industry datasets 250 free problems · No credit card See all Health & Insurance problems /problems/datasets/health