Inference hooks are in beta for Claude Enterprise, holding governed prompts for allow or deny before inference.

Policy gates move before model execution, so enterprise controls can stop prompts across Claude surfaces before inference.

For builders using claude.ai, Cowork, and Claude Code, governance now sits inline with signed requests and auditable denials.

• Aug 11, 2026ChatGPT can now show restaurant reservation times from OpenTable, Resy, and Yelp.
• Aug 11, 2026Claude Opus 5 launched with 1M context, 128k max output, and thinking on by default.
• Aug 11, 2026Anthropic removed modelcard text covering Claude Fable 5 and Claude Mythos 5.

This is the world we build in. It moves this fast. We move with it.