Our Preparedness Team will drive technical work, pushing the limits of our cutting edge models to run evaluations and closely monitor risks, including during training runs. Results will be synthesized in scorecards that track model risk.

→ View original post on X — @openai,