Claude Enterprise Adds "Inference Hooks" — Real-Time Allow/Deny Checks Before Every Prompt Reaches the Model

Anthropic has launched Inference hooks in beta for Claude Enterprise, letting organizations route every governed prompt through their own AI security server for an allow-or-deny verdict before inference runs, across claude.ai, Cowork, and Claude Code.