Track 2 · Building with models
Safety, Guardrails & Security
The model cannot touch your filesystem. Your harness can. This course teaches you to threat-model LLM applications, stop prompt injection, sandbox tools, keep secrets out of context, validate outputs, and build defense in depth around agents.
Course update · Jun 2026
OWASP LLM Top 10 references updated to the official 2025 edition (LLM01 through LLM10). Earlier drafts used 2023 numbering for several entries.