AI is transforming everything—from code generation to customer service—but many startups unknowingly build products that become exploitable attack vectors. As a Microsoft MVP in AI and global startup mentor, I’ll share real stories of how lean AI startups overlook model security, mishandle APIs, and leave dangerous gaps in LLM integrations. From prompt injection to over-permissive agents, this talk will dive into the intersection of AI innovation and exploitable design.

We'll also explore practical frameworks to help startups audit their AI pipelines—from open-source models to GenAI copilots—and mitigate risks before they scale. Attendees will walk away with actionable insights, red flags to watch for, and tools to strengthen AI security without killing startup velocity.

This is a talk for builders, hackers, and mentors who care about building smarter—and safer—AI.