Podcast

Data Privacy Playbook: Differential Privacy, Federated Learning, PETs & Consent UX

Open original DataTalks.Club episode

Data Privacy Playbook: Differential Privacy, Federated Learning, PETs & Consent UX

Original Episode

Use these links for the canonical episode and media sources.

• Open the original DataTalks.Club podcast page
• Watch on YouTube
• Listen on Spotify
• Listen on Apple Podcasts

Episode Overview

How can teams build useful machine learning while respecting user privacy, compliance, and re-identification risk? In this episode, Katharine Jarmul — privacy activist and Principal Data Scientist at ThoughtWorks Germany — walks through a practical Data Privacy Playbook focused on differential privacy, federated learning, privacy-enhancing technologies (PETs) and consent UX.

People

Use these links to connect the episode to guest notes.

• Katharine Jarmul

Chapter Summary

Use these checkpoints to decide whether to open the source transcript.

• 0:00 - Episode Introduction
• 1:40 - Guest Introduction: Katharine Jarmul — privacy activist, ML engineer, ThoughtWorks,
• 2:32 - Career Journey: data journalism, NLP, consulting, and machine learning
• 9:08 - Startup Focus: KI Protect, pseudonymisation, encrypted & federated ML
• 11:33 - Privacy Regulation Overview: GDPR, CCPA, CPRA and cookie consent defaults
• 14:35 - Cookie Consent & Opt-Out UX: one-click rejects and user behavior
• 16:24 - Defining Data Privacy: legal, social, and technical perspectives
• 21:35 - Practical Data Privacy (book): availability, previews, and giveaways
• 22:38 - Bridging Legal & Technical Views: privacy risk, translation, and collaboration
• 25:12 - User Profiling & Fingerprinting: browser history, apps, and re-identification
• 30:15 - Privacy-Friendly Personalization: session-based intent and ephemeral inference
• 33:08 - Privacy Engineering & PETs: encrypted ML, federated learning, and architecture
• 35:09 - Business Case for Privacy: risk management, regulation, and customer trust
• 40:50 - Differential Privacy Explained: formal definition, use cases, and libraries
• 45:08 - Anonymization Pitfalls: hashing, k-anonymity, Netflix de-anonymization lessons
• 47:00 - Designing for Privacy: consent, data minimization, and workflow practices
• 52:35 - Generative AI & Privacy: ChatGPT incidents, consent, retention, and enterprise
• 59:29 - Deploying Localized Models: Azure localization, fine-tuning, and ownership
• 1:01:15 - Further Learning: Probably Private newsletter, notebooks, and differential