Frontier Safety Framework — DeepMind
Risk thresholds, capability evals, and escalation policies for frontier models — safety as an operational protocol, not a slogan.
Google DeepMind · Google DeepMind
Framework documents matter because they bind capability claims to concrete eval gates and misuse categories.
Headtap tracks whether labs update thresholds when models cross them — the gap between paper and practice.
Primary source
Read the original: Google DeepMind — Frontier Safety Framework — DeepMind