IDC PeerScape: Practices for Securing AI Models and Applications

This IDC PeerScape describes the best practices for securing AI models and applications."Cybersecurity vendors are securing their AI applications and models by protecting APIs, monitoring model inputs and outputs, and proactively looking for weaknesses," said Michelle Abraham, research director, Security and Trust at IDC. "They have well-thought-out protections in place using existing security technologies as well as new technologies designed for GenAI."

IDC PeerScape Figure

Executive Summary

Peer Insights

• Practice 1: Protect APIs and Connections to AI Infrastructure as Well as the AI Infrastructure Itself
  • Challenge
  • Examples
    • Broadcom
    • CrowdStrike
    • IBM
    • Trend Micro
  • Guidance
• Practice 2: Use Verified and Tested Foundation Models
  • Challenge
  • Examples
    • Broadcom
    • Cisco
    • CrowdStrike
    • IBM
    • Trend Micro
  • Guidance
• Practice 3: Monitor Model Inputs and Outputs to Detect and Respond to Attacks Against AI
  • Challenge
  • Examples
    • IBM
    • Cisco
    • Splunk
  • Guidance