Implementing AI While Respecting Privacy, Impossible?
summary
An engineer from Proton describes how they built privacy-preserving AI features by running a local LLM inside the browser and also via a server-side option. The talk covers technical challenges of making a browser-based LLM work alongside end-to-end encryption and other security/privacy constraints, plus practical trade-offs like bandwidth costs and related ethical questions.
AI-written summary · about
chapters (11)
- 0:00Introduction
- 4:41Privacy vs generative AI
- 8:54Local solution no trade-offs
- 10:16Major technical constraints
- 15:18Browser and hardware compatibility
- 19:33Security and CSP architecture
- 22:01UX in the composer
- 24:50Server-side alternative
- 27:42Simplified frontend refactor
- 29:04Conclusion and future outlook
- 31:54Q&A
related talks
06mData Privacy Concerns and Solutions When using LLMs in Production
28mWhy Private AI Matters: Data Sovereignty as the Foundation for Trustworthy AI on the Open Web
46mThe Ultimate Guide to Browser-Based AI: Privacy-First, Offline-Ready and Future-Proof
19mPrivate AI: Building Trust Through Verifiable Computation
33mLocal Models: How to Run AI in the Browser
14mToward Provably Private Insights into AI Use
41mEscaping the Cloud: High-Performance AI in your Browser
36mData privacy in an AI-driven world