AI Application Security for Developers: Secure LLMs, RAG, and Agents
Course overview
Design and assess defensive controls for AI-enabled applications. Build threat models, enforce authorization outside the model, protect retrieval data, and test agent permissions using isolated, authorized labs.
For experienced developers. Estimated total study time: 24 hours including practice and capstone work; not a claim of recorded video duration.
Learning outline
- Threat modeling and trust boundaries (3h)
- Prompt injection and untrusted content defenses (4h)
- Secure RAG and tenant-aware retrieval (4h)
- Least-privilege tools and human approval (4h)
- Output validation, secrets, and monitoring (4h)
- Capstone: security review and regression tests (5h)
Prerequisites
Experience building APIs; basic LLM/RAG knowledge; Git and a local development environment. Security exercises only on owned or explicitly authorized systems. Use synthetic data and restricted test credentials.
Development status
Draft course plan. Lessons, recordings, downloadable materials, and assessments are not yet uploaded. Tuition is awaiting approval; any cloud or third-party subscription charges are separate.
