As an AI startup, you're under pressure: ship fast, iterate faster and security may end up at the bottom of the list. But as AI agents become a core part of your product, your attack surface grows significantly, and traditional security mechanisms are no longer enough. Threats like prompt injection and excessive agency can cost you customers, trust, and ultimately your business.

This talk presents research-based, practical tools for building security in from day one. It draws on the whitepaper from the research project Security by Design for AI Startups, led by the Alexandra Institute.