BARC-Bench shows frontier agents still breach unstated workplace boundaries on benign asks
Even the strongest tested workplace agent still violates soft boundaries on more than a quarter of benign admin tasks.
Read the piece#safety
Published Frontier reporting tagged “safety”.
2 articles · Full archive · Topics
Stories on safety
1 recent piece
More tagged safety
1 story
OpenAI releases MentalHealthBench: 1,215 expert-rubric chats, with GPT-6 Astra top at 57.3
OpenAI built the open benchmark with more than 80 licensed clinicians. It scores the next reply in synthetic chats that range from everyday stress to emergencies, and an OpenAI model does the grading.
One brief a week.
New releases, then the constraint.