AI is moving from tool to autonomous actor, and most organizations are still treating it like software.
In this episode, Kathryn Wang, Principal Public Sector at SandboxAQ, breaks down what actually changes when AI systems move into production, why security models are falling behind, and how the real battleground is shifting toward the data layer.
The conversation explores how agentic AI introduces entirely new threat vectors, why identity and authorization are becoming the primary attack surface, and how quantum computing will reshape encryption, national security, and enterprise risk.
For leaders, the takeaway is simple but uncomfortable: this is no longer about adopting AI faster. It’s about understanding what you’re exposing before it’s too late.
⸻
👤 About the Guest
Kathryn Wang is Principal, Public Sector at SandboxAQ, working at the intersection of AI, cybersecurity, and quantum technologies.
She previously spent over two decades at Google, where she worked across product, strategy, and innovation, including early-stage AI initiatives.
Her work today focuses on helping governments and enterprises navigate emerging risks in AI systems, data security, and post-quantum cryptography.
https://www.linkedin.com/in/kathryn-wang/
⸻
🔑 Key Takeaways
• AI is no longer just generating content, it is executing actions within systems
• Authorization is becoming the biggest security risk in the age of agentic AI
• Most organizations still treat AI as a tool, not as an autonomous actor
• Data is the ultimate target, whether customer data, IP, or AI training data
• Quantum computing will redefine encryption and expose weak cryptographic systems
• Sovereign AI is emerging, shaped by national values, policies, and data control
• Human oversight alone is no longer enough to manage AI-driven systems
• Security needs to shift from layered defense to protecting the data layer itself
⸻
🎯 What You’ll Learn
• What fundamentally changes when AI moves from research to production
• Why agentic AI creates new attack surfaces that traditional security cannot handle
• The biggest AI risks organizations are underestimating today
• How AI can be weaponized through authorized systems and workflows
• Why securing the data layer is more important than adding more security tools
• How quantum computing impacts cybersecurity, banking, and national security
• What sovereign AI means and how it will shape global technology competition
⸻
⏱️ Episode Highlights
00:00 Introduction and Kathryn’s journey from Google to SandboxAQ
03:00 What changes when AI moves into production environments
07:30 The most underestimated AI risks in organizations today
12:00 Agentic AI, authorization, and new threat models
16:00 Why the data layer is the real battleground
22:00 Is cybersecurity still reactive in the AI era
27:00 Sovereign AI and global competition dynamics
32:00 Governance, liability, and who is responsible for AI decisions
37:00 Quantum computing and the future of encryption
43:00 Why IP is data and must be secured at all costs
45:00 Final thoughts and practical ways to learn AI
⸻
📚 Resources Mentioned
• SandboxAQ: https://www.sandboxaq.com/
• LinkedIn for AI and cybersecurity learning
• NotebookLM for simplifying complex topics
