Managing Risk and Security for GenAI and PredAI (On-Demand)

Course Description

Use of Generative AI (GenAI) has increased dramatically over the last two years, and companies by and large are struggling to get a handle on risk management on both GenAI and Predictive AI (PredAI). While legal obligations are one lens through which to establish a risk management program for use of AI, such a program would not be sufficient to manage the associated risks. Join this session to learn about the risks associated with GenAI and PredAI and compliance obligations. In particular, this session will review the security risks associated with adversarial attacks on AI, including evasion, poisoning, privacy and abuse attacks.

Principles

• Generative AI vs. Predictive AI
• Overview and update on AI legislative frameworks
• Risk management frameworks
• Security risks for GenAI and PredAI

Syllabus

1. Understanding Generative AI vs. Predictive AI
   • What is GenAI vs. PredAI?
   • Is focusing on PredAI necessary?
2. Overview and update on AI legislative frameworks
   • US legal update
   • EU legal update
   • Other international developments
3. Risk management frameworks
   • NIST AI RMF
   • ISO, IEEE, Palo Alto and other frameworks
4. Security risks for GenAI and PredAI
   • Attack classifications (direct vs. indirect, attacker knowledge)
   • Types of attacks (evasion, privacy, poisoning and abuse)
   • Mitigations