Added `overview` as a proper entity type throughout the system, migrated all 36 overview pages to `entityType: overview`, built overview-specific InfoBox rendering with child page links, created an OverviewBanner component, and added a knowledge-base-overview page template to Crux.
Policy approaches establish organizational frameworks for responsible AI development.
Scaling Policies:
Model Documentation:
Evaluation Governance:
Development Speed:
Proposals to pause or slow frontier AI development until safety is better understood, offering potentially high safety benefits if implemented but ...
Evals-based deployment gates require AI models to pass safety evaluations before deployment or capability scaling.
Responsible Scaling Policies (RSPs) are voluntary commitments by AI labs to pause scaling when capability or safety thresholds are crossed.
Model specifications are explicit written documents defining desired AI behavior, values, and boundaries.