InfosecGirls Virtual Meet - 17th May, 2025

May 17, 2025

InfosecGirls Session - 17th May, 2025

 

Topic: LLM security for builders — prompt injection basics, data leakage risks, and guardrails when shipping AI-assisted features.

 

Summary

  • Threat overview for apps that call external or hosted language models.
  • Prompt injection and indirect injection in RAG and tool-calling flows.
  • Data handling: minimising sensitive context in prompts and logging redaction.
  • Practical guardrails: output policies, human review hooks, and abuse monitoring.