Companies deploying AI models often lack expertise in security hardening specific to machine learning systems.
Provide audits that test for model poisoning, prompt injection, data leakage, and other AI-specific threats.
Target both AI vendors and enterprises implementing solutions, charging per audit.
Start with testing common vulnerabilities in popular deployment methods.
Barrier is establishing credibility in this emerging field.