In 2025, secure AI has evolved from blocking generative AI at firewalls to managing production-scale deployments with business-critical data. Organizations now navigate a complex landscape where traditional application security mistakes resurface alongside novel AI-specific risks. The shift from experimental pilots to regulated production environments demands comprehensive governance across four critical layers: model, application, infrastructure and data security. Shadow AI persists, agentic AI experimentation raises authentication and accountability questions, and adversarial misuse remains real but not yet game-changing.

In this insightful session, Anton Chuvakin, senior staff security consultant, office of the CISO, Google Cloud, will share insights on:

• Why securing all four critical layers is non-negotiable;
• Practical AI use cases delivering real value, from auxiliary support to breakthrough capabilities;
• The governance challenges and resilience considerations shaping AI security strategies.