Broad access through consistent standards

OpenAI published a safety alignment plan requiring consistent safeguards whether models are accessed directly, through hosting partners, or resellers. Updated Usage Policies take effect November 20, 2026. Model routers and AI gateways must obtain prior OpenAI review and approval. The plan introduces SafetyIDs, letting customers provide identifiers so OpenAI can distinguish end users in abuse investigations without knowing real identities. Enforcement is framed as evidence-based and proportional.