Common Elements of Frontier AI Safety Policies

Common Elements of Frontier AI Safety Policies

Why it matters

Laid out 9 common elements across 12 companies' policies and became a de facto quasi-standard document for the industry

Source — Primary sources first; secondary reports are labelled “Press”

Officialmetr.orghttps://metr.org/common-elements

Added in Week 35, 2026 · See that week

Same week

Kakao open-sources Kanana-2 (Base/Instruct/Thinking)OfficialYann LeCun confirms he has left Meta to found AMI LabsPressAnthropic publishes Agent Skills as an open standard (agentskills.io)PressNVIDIA applies VLMs and VFMs to semiconductor defect classificationOfficial

Nearby on the same axis

xAI publishes its Frontier Artificial Intelligence FrameworkOfficialAnthropic reports prompt injection defense results (Claude for Chrome)OfficialNatural Emergent Misalignment from Reward HackingPaperCalifornia AG opens investigation into xAI/Grok non-consensual sexual deepfakesGovernment