Apple Workshop on Privacy-Preserving Machine Learning & AI 2026

2026-05-08 · Source: Apple Machine Learning Research · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Cybersecurity & Data Privacy · Depth: Expert, short

Summary

Apple recently hosted a two-day Workshop on Privacy-Preserving Machine Learning & AI, bringing together Apple researchers and the broader academic community. The event focused on three core areas: Private Learning and Statistics, Foundation Models and Privacy, and Attacks and Security. Discussions covered advances and open questions in privacy and ML, including federated learning, statistical learning, trust models, attacks, privacy accounting, and challenges specific to foundation models. The workshop featured presentations such as "Crypto for DP and DP for Crypto" by Kunal Talwar and "Understanding and Mitigating Memorization in Foundation Models" by Franziska Boenisch. Additionally, 23 published works were presented, exploring topics like adaptive methods in high privacy settings, memorization in Clip and Diffusion models, combining ML with homomorphic encryption, and efficient privacy loss accounting.

Key takeaway

For research scientists developing AI systems, understanding the latest in privacy-preserving ML is critical. You should explore techniques like federated learning, differential privacy, and homomorphic encryption to ensure user data protection, especially when working with large foundation models. Prioritize rigorous security evaluations to bridge theoretical privacy frameworks with practical, real-world applications.

Key insights

Privacy-preserving AI research is crucial for integrating advanced AI capabilities while safeguarding user data.

Principles

• Privacy is a fundamental human right.
• Rigorous evaluation grounds innovation.
• Bridging theory with real-world applications.

Method

The workshop explored privacy-preserving ML through federated learning, statistical learning, trust models, attack analysis, and privacy accounting, particularly for foundation models.

In practice

• Mitigate memorization in foundation models.
• Apply homomorphic encryption in ML.
• Use adaptive methods in high privacy settings.

Topics

• Privacy-Preserving Machine Learning
• Differential Privacy
• Foundation Models
• Federated Learning
• AI Memorization

Best for: Research Scientist, AI Scientist, AI Security Engineer, Machine Learning Engineer

Related on AIssential

• Secure and Privacy-Preserving Vertical Federated Learning — A novel framework enhances vertical federated learning privacy via distributed aggregation and secure multiparty computation.
• Presentation: Busting AI Myths and Embracing Realities in Privacy & Security — Effective AI privacy and security demand proactive, continuous organizational effort, not reliance on external fixes or single-solution myths.
• Silent Failures in Federated Personalization of Foundation Models — Federated personalization of foundation models creates "Silent Failures" like bias and fairness collapse, undetectable due to privacy constraints.
• How probability models protect privacy — Differential privacy uses calibrated randomness to protect individual data while enabling accurate aggregate analysis.
• Is there a notable increase in demand for privacy-preserving AI/ML with the advent of LLMs? [D] — Despite performance trade-offs, regulatory pressure and increasing data privacy risks are driving demand for privacy-preserving AI.
• Benchmarking Empirical Privacy Protection for Adaptations of Large Language Models — Distribution shifts critically impact practical privacy in differentially private LLM adaptations, often undermining theoretical guarantees.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Apple Machine Learning Research.