The Prompt Guardians: Threats and Security Recipes for Zen Production
summary
A practical presentation on securing production GenAI applications, covering threats like prompt injection and data leakage and regulatory constraints (EU AI Act, NIST). It shows defensive architecture and observability using tools such as LangSmith and Langfuse, I/O AI guards, semantic filtering, policy-as-code, quotas/rate limiting, continuous security QA with CI/CD attack suites (Garak) and external pentests, and provides a concrete playbook to move from demo to hardened production.
AI-written summary · about
chapters (11)
- 0:00AI app security intro
- 1:03Gen AI at Moodis
- 3:49Financial risks in chatbots
- 7:02EU AI regulation pressures
- 10:09Hotel chatbot vulnerability demo
- 15:00Prompt injection techniques
- 22:07Securing AI apps best practices
- 26:16AI flow observability
- 29:26Guardrails implementation
- 39:02Continuous security testing
- 41:50Conclusion
related talks
39mSecuring AI at Scale: Practical Defenses against Prompt...
18mCommit Secure AI: Embedding Security in Code, Pipelines, and Production
26mSponsored Session: Navigating Open Source AI with Guardrails
35mGenAI Application Security: Not Just Prompt Injection
53mHow to Break AI Systems (Before Someone Else Does)
21mProtect your generative AI applications against jailbreaks
45m📙 Prompt, Commit, Repeat: Security at Scale When 1,000 Devs Go AI-Native
09mSecuring Generative AI Workloads