Still, the message is clear: Current AI language models accept fabricated medical statements at rates that should concern anyone relying on them for health guidance. Even GPT-4o, the strongest ...