ELLIS Workshop on Safe and Secure Artificial Intelligence

The objective of this online workshop is to discuss recent scientific advances in the areas of privacy, safety, trust, biases, and interpretability in artificial intelligence.

For all information, the agenda and to register to receive the Zoom link, visit the workshop webpage: https://ellis.ciirc.cvut.cz/ellis-workshop-on-safe-and-secure-ai-2026/

Organizers

The workshop is co-organized by the ELLIS Unit Prague, ELLIS Associate Unit Lviv, the ELLIS Unit Alicante, and the ELLIS Program on Human-Centric Machine Learning.

For questions, please feel free to reach to:

Rostyslav Hryniv or Tetiana Zakharchenko, ELLIS Associate Unit Lviv
Nuria Oliver, ELLIS Unit Alicante and ELLIS Program on Human-Centric Machine Learning
Josef Šivic, ELLIS Unit Prague

Agenda

This event will take place online on the 20th of May, 2026. Times are in CEST (UTC+2).

Time	Description
13:00-14:30	Session I Nuria Oliver – What foundation models learn from us and use it against us Mykola Pechenizkiy – When Benchmarks Mislead: Rethinking AI Safety Evaluation Novi Quadrianto – Performative prediction made practical
14:30-15:00	Coffee Break
15:00-16:30	Session II Krishna Gummadi – In Agents We Trust, but Who do Agents Trust? On latent preferences and perceptions that steer LLM generations Lesia Semenova – From Model Uncertainty to Trust: Safe and Interpretable AI under the Rashomon Effect Valentyn Melnychuk – Causal Machine Learning under Privacy Constraints (short talk) Denys Herasymuk – Smarter Together, Safer Apart: Achieving Privacy and Robustness in Personalized Machine Learning (short talk)

Live-stream

The live event will be streamed on YouTube:

Ask your questions via SLIDO: #ellis-2026 or simply scan this QR code:

Experiment on the homogenization of aesthetics