ChatGPT Health Misdiagnoses Half of Emergencies, Study Finds

A new study revealing ChatGPT Health's failure to correctly triage half of medical emergencies marks a significant inflection point for AI in medicine. Published in Nature Medicine, the results challenge the narrative of generalist models as cure-alls for specialized, high-stakes applications. This isn't just a technical stumble; it’s a strategic setback for the broader industry push to deploy AI agents in critical, consumer-facing roles, questioning their fundamental readiness for real-world clinical environments. This outcome immediately benefits more specialized, clinically-validated health tech incumbents and puts immense pressure on Big Tech to prove the safety of their models. The findings could trigger a new wave of regulatory scrutiny from bodies like the FDA, creating a crisis of confidence among healthcare providers. This may slow enterprise adoption in risk-averse hospital systems and force a strategic retreat towards lower-stakes wellness applications until model reliability can be guaranteed and proven.