Anthropic - Claude Certified Developer - Foundations Study Guide

7: Security and Safety

This chapter hardens Claude applications against adversarial input, data loss, and misconfigured trust. It covers prompt injection and jailbreak defense, untrusted-input handling, PII and data-leakage controls, the layered guardrail model, hooks as deterministic safety gates, and the identity, secrets, and key-management practices that separate a prototype from a production deployment.