15 frontier LLMs: 94% of correct medical answers fail under adversarial pressure
davidmanheim · x · 2026-09-06
Discussing a study of 15 frontier LLMs in healthcare, the author raises a counterpoint: doctors already face adversarial pressures from pharma sales reps and insurance rules that degrade care, while AI is immune to those — so are human adversarial risks worse than AI's? Context: models score 80%+ on static exams like MedQA, but a median 94% of previously correct answers failed under adaptive adversarial testing.
More from Safety
- LBC interview discusses OpenAI, rogue AI agents, and AI accountability gaps — ShakeelHashim · 2026-09-06
- AI risk discourse has made bioterrorism and cyber threats sound like obvious near-term dangers — EigenGender · 2026-09-06
- GPT-6 reportedly jailbroken via extended task-in-prompt attack — EducationalCicada · 2026-09-06
- xAI fails to block Minnesota's AI nudification ban; lawsuit continues — VraserX · 2026-09-06
- Google's always-on Gemini Spark agent handles photos and trips, raising fresh privacy questions — emmanuelvivier · 2026-09-06
- Los Angeles school district bans generative AI in schools for one year — emmanuelvivier · 2026-09-06