Hallucinations, Prompt Manipulations, and Mitigating Risk: Putting Guardrails around your LLM-Powered Applications

Learn how to mitigate risks in LLM-powered applications through effective guardrail strategies in this 32-minute conference talk presented by Don Shin of CrossComm at All Things Open AI 2025. Discover pre-processing techniques to protect against end-user prompt manipulation, approaches for evaluating LLM outputs using LLM-as-a-judge, and an overview of open source guardrail frameworks. The presentation addresses how enterprises can overcome adoption barriers caused by hallucinations - where LLMs respond inaccurately but with complete confidence - and includes a live demonstration of implementing these risk mitigation strategies in real-world production settings.