When Data Can’t Move but Insight Must
summary
When data must stay put, privacy-enhancing technologies actually move the insight: deployments use homomorphic encryption and secure multi-party computation to run joint analyses without raw-data transfers, and differential privacy or synthetic data to release safe outputs. National pilots combine hospital, insurer and supermarket records using MPC/HomE with computing nodes at the statistical office; success comes from clear objectives, contractual smart‑contracts, and vendor maturity. Risks include LLM-enabled reidentification, limited query budgets, and heavy engineering plus legal coordination; many libraries are open source and audited.
AI-written summary · about
chapters (16)
- 0:00Privacy enhancing technologies
- 7:19Why PETs matter
- 15:28PETs in official statistics
- 23:21PETs for generative AI
- 29:44Data governance with PETs
- 36:32Industry PETs use cases
- 42:18PETs success in statistics
- 45:53Adopting PETs organizationally
- 48:59PETs transparency challenges
- 55:20Monitoring data quality
- 58:31Digital sovereignty with PETs
- 1:00:38PETs impact on AI
- 1:05:13PETs regulations
- 1:10:49Future of PETs standardization
- 1:16:08Open source PETs
- 1:18:12Closing remarks
related talks
26mDoes using data mean giving up privacy?
26mAnalytics & Data Science with Privacy at its Core
01h 59mDeep Dive Workshop – Data Security and Data Protection
20mShare data without revealing personal information
41mMatching Uses and Data Protections for Government Data
28mInsights, Not Secrets: How Synthetic Data Empowers Privacy
29mIntro to Privacy-Enhancing Technologies (PETs)
29mPrivacy-preserving methods: Building secure projects