← Browse

World's Fair 2024

LLM Safeguards: Security Privacy Compliance Anti Hallucination: Daniel Whitenack

Overview

This talk addresses the practical challenges of deploying secure, accurate, and compliant AI systems, particularly focusing on open-access large language models (LLMs) in enterprise environments. It outlines common risks such as hallucinations, supply chain vulnerabilities, data breaches, and prompt injection, and proposes mitigation strategies. The core thesis is that by understanding and proactively addressing these risks with existing and emerging technologies, organizations can build more trustworthy AI applications.

Who should watch

Key takeaways

Notable quotes

*The text on the internet sometimes represents accuracy and not or it might have outdated information or it could just spew out.*
*There's definitely a challenge here with confident answering of inaccurate information.*
*This is a new way that that enters in so we've got these supply chain vulnerabilities.*

Watch on YouTube →

Unofficial community note. Prefer the recording for nuance.