Anthropic adds security-posture review, inference hooks and administrative controls.
Security teams gain more ways to inspect configuration and apply policy to agent activity.
What happened
Anthropic’s 5 November update describes a security centre in beta, generally available Inference: Running a trained model to get answers, as opposed to training it. Glossary hooks for prompts and pre-tool calls, centralised management in beta, and additional access controls. Availability varies by feature; the security centre is not generally available simply because inference hooks are.
Why it matters: Security teams gain more ways to inspect configuration and apply policy to agent activity.
Sources: Primary source, “More visibility and control over Claude Enterprise for security teams - Claude by Anthropic”; Primary source, “Review your organization's security posture with the security center - Claude Help Center”; Primary source, “Inference hooks overview - Claude Help Center”