Security and Safety

Secure keys, defend against prompt injection, filter content and respect the usage policy.

1. 7.1API Key and Secret SecurityWhere API keys may and may not live, how to scope and rotate them with workspaces and spend limits, and what to do the moment a key leaks.8 min
2. 7.2Prompt Injection and Untrusted InputWhat prompt injection is, how it reaches an application through documents, web pages and tool results, and the layered defences developers are expected to build.9 min
3. 7.3Content Moderation and GuardrailsUsing Claude as a moderation classifier, filtering inputs and outputs, handling refusals correctly, and designing guardrails that stay inside your application rather than relying on the model alone.8 min
4. 7.4Usage Policy and Safe DeploymentWhat Anthropic's Usage Policy requires of developers, which deployments carry extra obligations, and the practical steps that make a Claude-powered product responsible by default.8 min