OpenAI lawsuit and ChatGPT Health rollout under scrutiny

The civil suit filed by Florida pastor Scott Winters on July 21 2026 alleges that ChatGPT‑4o’s advice to stay home after reporting dizziness and groin pain contributed to a pulmonary embolism, prompting a request for damages and an injunction to halt the newly launched ChatGPT Health feature until independent safety certification is obtained. OpenAI maintains that the chatbot is a general‑information tool, not a medical device, and its terms direct users to seek professional care for urgent issues.

OpenAI completed a nationwide rollout of “Health in ChatGPT” on July 23 2026, allowing U.S. adults to link Apple Health, electronic‑medical‑record systems and other wellness apps so the model can reference personal data when answering health queries. The service runs on GPT‑5.6 Sol for paid tiers and GPT‑5.5 Instant for free users, with OpenAI reporting performance on its internal HealthBench benchmark at or above clinician levels. Weekly health queries now exceed 300 million worldwide.

By July 30 2026, U.S. healthcare professionals warned that patients might treat AI‑generated responses as medical advice, raising privacy, consent, data‑security and patient‑provider relationship concerns. Simultaneously, scientific studies offered conflicting assessments of AI performance: a Nature Medicine paper found general‑purpose models (ChatGPT, Claude, Gemini) outperformed a specialized model (OpenEvidence), while a later JAMA study rated OpenEvidence higher on accuracy, utility and verifiability. The divergent findings underscore calls for transparent, independent evaluation of AI health tools before broader clinical adoption.