AI + a16z

Security Founders Talk Shop About Generative AI

May 15, 2024

Security-startup founders Dean De Beer, Kevin Tian, and Travis McPeak discuss generative AI, AI models for security, phishing attacks, SEO poisoning, and utilizing generative AI for enhanced security practices. They highlight the importance of accuracy, repeatability, user experience, context, intent, data enrichment, and tackling alert fatigue in security operations.

22:31

Creator website

Episode guests

Dean De Beer

Kevin Tian

Travis McPeak

AI Summary

AI Chapters

Episode notes

Podcast summary created with Snipd AI

Quick takeaways

Customizing large language models boosts accuracy in AI security applications.

Balancing model selection for security needs considers hosting, inference time, and data handling.

Deep dives

The Importance of Iteration in AI Security Prototypes

Building secure AI prototypes involves more than initial success. Dean from Command Zero emphasizes the importance of continual accuracy in AI security. He discusses the challenges of repetitive accuracy, avoiding data exclusion, and ensuring secure automation. Dean highlights the necessity of thoroughness and consistency in deploying large language models for security products.

Introduction

2min

Challenges and Considerations in AI Model Selection for Security Applications

10min

Exploring Scammers' Targeting of AI Users for API Keys through Phishing Attacks

2min

A Detailed Exploration of SEO Poisoning and Social Engineering Tactics

2min

Embracing Generative AI for Enhanced Security Practices

7min

In this bonus episode, recorded live at our San Francisco office, security-startup founders Dean De Beer (Command Zero), Kevin Tian (Doppel), and Travis McPeak (Resourcely) share their thoughts on generative AI, as well as their experiences building with LLMs and dealing with LLM-based threats.

Here's a sample of what Dean had to say about the myriad considerations when choosing, and operating, a large language model:

"The more advanced your use case is, the more requirements you have, the more data you attach to it, the more complex your prompts — ll this is going to change your inference time.

"I liken this to perceived waiting time for an elevator. There's data scientists at places like Otis that actually work on that problem. You know, no one wants to wait 45 seconds for an elevator, but taking the stairs will take them half an hour if they're going to the top floor of . . . something. Same thing here: If I can generate an outcome in 90 seconds, it's still too long from the user's perspective, even if them building out and figuring out the data and building that report [would have] took them four hours . . . two days."

Follow everyone:

Check out everything a16z is doing with artificial intelligence here, including articles, projects, and more podcasts.