AI Safety and Security

The OWASP LLM Top 10, prompt-injection offense and defense, guardrails, and red-teaming your own app. Ada's differentiator.

1. 1The LLM attack surface and the OWASP LLM Top 103 q
2. 2Prompt injection: direct (LLM01)3 q
3. 3Indirect injection and RAG poisoning (LLM08)3 q
4. 4System-prompt extraction and leakage (LLM07)3 q
5. 5Guardrails and input/output filtering3 q
6. 6Tool and agent abuse prevention (LLM06)3 q
7. 7Sensitive-data disclosure and exfiltration (LLM02)3 q
8. 8Red-team your own app: the attack harness3 q