News

Amartya has been awarded DKK 7,198,686 for Data-Adaptive Privacy for Machine Learning at the University of Copenhagen. The grant comes through Independent Research Fund Denmark (DFF)’s thematic call on research in artificial intelligence (2026–2029).

Amartya has become an ELLIS Scholar!

Congratulations to Johanna, Vikrant, and Christian on their NeurIPS 2026 acceptances! Johanna's paper on adaptive private k-PCA was selected as a spotlight. Vikrant's paper on privately estimating monotone statistics and Christian's Lumberjack paper on differentially private random forests were also accepted.

Our paper on private question answering without private training was accepted to InfPriv: The New Landscape of AI Privacy, a workshop at NeurIPS 2026.

Amartya co-organised the Learning Theory Summer School and Workshop and the Workshop on Differential Privacy and Unlearning in Machine Learning in Copenhagen.

Giorgio's paper on learning-theoretic foundations for model collapse was accepted to ICML 2026. Luka and Vikrant's paper on why forget-only unlearning needs memorization was accepted to the Workshop on the Impact of Memorization on Trustworthy Foundation Models at ICML 2026.

Three papers from the group were accepted to TPDP 2026 (Theory and Practice of Differential Privacy): Johanna and Yaxi's paper on LoRA and privacy, Max's paper on differentially private fair learning, and Carolin's paper on the right sensitivity notion for unlearning.

We have three new members joining our group. Luka and Giorgio have joined us since September 1 as PhD students and Vikrant joined us from January 1, 2026 as fixed term Assistant Professor.

Johanna's paper on Differentially Private PCA was accepted to NeurIPS 2025 and Carolin's paper on Online Learning in an Echo Chamber was accepted to SODA 2025 and selected for FORC (Foundations of Responsible Computing) highlights.

Amartya is the co-director of the P1 program on Data Privacy in Machine learning!