AI Without Access: Shipping Intelligence When You Can’t See User Data

Everyone is racing to plug AI into their workflows. But what happens when your product is built on a promise to never see user data in the first place? As CTO of a zero knowledge product, Frederic Rivain has to reconcile two strong forces: teams who want to use AI everywhere, and an architecture that is designed to reveal nothing.
In this talk, Frederic unpacks how Dashlane uses AI internally in a secure and responsible way, and how they choose and integrate AI into the product while keeping their zero knowledge model intact. He will share the trade offs, the architectural patterns, and the guardrails that let teams experiment with AI without weakening the trust the product is built on.
Challenges:

• Leveraging AI internally when sensitive customer and company data must remain encrypted or out of scope
• Choosing AI providers and architectures that align with strict privacy, security and compliance requirements, not just convenience
• Embedding AI into critical user journeys without breaking the zero knowledge promise or eroding customer trust
  Key Learnings:
• How to design AI use cases around a zero knowledge mindset, including what to avoid, what to transform, and where AI genuinely adds value
• Practical patterns for using AI internally while protecting data, from governance and access controls to what you never put in a prompt
• A decision framework for selecting and integrating AI partners in high trust products so that privacy and security stay first class, not an afterthought