Apple Releases Recordings and Research From Privacy-Preserving AI Workshop

Apple has released four session recordings and a comprehensive research recap from its 2026 Workshop on Privacy-Preserving Machine Learning & AI. The disclosure, published on May 11, 2026, provides insight into how the company is addressing the intersection of artificial intelligence and data security.

The two-day event brought together Apple researchers and members of the broader academic community to discuss the latest developments in privacy-preserving machine learning (ML) and AI. The workshop focused on three primary pillars: Private Learning and Statistics, Foundation Models and Privacy, and Attacks and Security.

According to Apple, the discussions explored various advances and open questions regarding the implementation of AI that protects user data. This included a focus on federated learning, a method where AI models are trained across multiple decentralized devices without exchanging the raw data itself, as well as statistical learning, trust models, and privacy accounting.

The company also addressed the specific challenges posed by foundation models—large-scale AI systems trained on vast datasets that can be adapted for many different tasks—and how to maintain privacy within those architectures.

Presentations and discussions at the workshop explored advances and open questions in privacy and ML, including federated learning, statistical learning, trust models, attacks, privacy accounting, and the unique challenges presented by foundation models. These research areas ground innovation in rigorous privacy and security evaluation, bridging theoretical frameworks with real-world applications.

Apple

As part of the release, Apple highlighted four specific talks that illustrate the technical breadth of the workshop. One featured presentation, Crypto for DP and DP for Crypto, was delivered by Apple Research Scientist Kunal Talwar. This talk focused on the relationship between cryptography and differential privacy (DP), a system that allows for the analysis of patterns in a dataset while ensuring that individual identities remain protected.

Other featured presentations included:

• Online Matrix Factorization and Online Query Release, presented by Aleksandar Nikolov of the University of Toronto.
• Learning from the People: Communicating about S&P Technology for Responsible Data Collection, presented by Elissa Redmiles of Georgetown.
• Understanding and Mitigating Memorization in Foundation Models, presented by Franziska Boenisch of CISPA.

Beyond the recorded sessions, Apple noted that 24 published works were presented during the workshop. Among these were three papers developed by current and former researchers at Apple, including research focused on combining machine learning with homomorphic encryption, a form of encryption that allows computations to be performed on encrypted data without needing to decrypt it first.

The release of these materials follows Apple’s ongoing efforts to publish its machine learning research and provide transparency into the theoretical frameworks it uses to evaluate the security of its AI implementations.