Security teams now rely on AI for detection, triage, and response. Few can show evidence that this trust is earned. Policies say "use AI responsibly," but auditors and incident reviewers want proof, and most teams can't explain why their AI flagged one alert and missed another.

This session turns AI trustworthiness into something a security program can measure and govern. We'll cover what to test before deployment (accuracy on your own data, drift, adversarial robustness), how to trace automated decisions back to their inputs, and how to document them so they hold up in an audit or post-incident review. We'll map these practices to NIST AI RMF, ISO 42001, and GDPR expectations, then look at where governance breaks down: unowned models, shadow AI, vendor black boxes, and checkbox reviews.

Takeaways: a trust scorecard for AI security tools; a model inventory and ownership template; an explainability checklist for vendors; metrics that show whether governance works.