Academy / Responsible AI & Governance

Path 4 · Course 4 · For security, technology, product and risk teams

AI Security & Safety

Prompt injection, jailbreaks, data poisoning, model theft and supply-chain attacks; the OWASP Top 10 for LLM applications; securing AI agents; red-teaming and frontier AI safety.

📘 6 modules⏱ About 3.5 hours✅ Knowledge check per module🎓 Final assessment & certificate🎯 Some technical familiarity helps

0 of 6 modules complete

Start Module 1 Final assessment

Modules

  1. 1

    The AI threat landscape

    Why AI security is different, AI as target, tool and weapon, safety vs security, key frameworks and the AWS scoping matrix.

    25 min
  2. 2

    Prompt injection and jailbreaks

    Direct and indirect prompt injection, jailbreak techniques, why there's no complete fix, and defence in depth.

    30 min
  3. 3

    Attacks on data and models

    Data poisoning and backdoors, model extraction and theft, privacy attacks, adversarial examples and AI supply-chain risk.

    30 min
  4. 4

    Securing LLM applications

    The OWASP Top 10 for LLM Applications 2025: disclosure, output handling, prompt leakage, RAG weaknesses, misinformation and consumption.

    30 min
  5. 5

    Securing AI agents

    Why agents raise the stakes, excessive agency, the dangerous combination, agent controls, MCP and tool risks, and browser agents.

    30 min
  6. 6

    Red teaming, evaluation and secure operations

    AI red teaming and Microsoft's lessons, running an exercise, safety evaluation, frontier AI safety, secure operations and a checklist.

    30 min