2025: Year Dominated By AI

This Top 10 is based on statistics relating to visits to our news section, media coverage, and the resonance of our topics on social media. We have weighted these figures to ensure that all EPFL faculties are represented. AI dominates the list, and we have not included all articles related to Apertus in this Top 10.

• Switzerland at the forefront of AI

EPFL, ETH Zurich and the Swiss National Supercomputing Centre (CSCS) have launched Apertus, Switzerland's first large-scale, open, multilingual language model - a milestone in generative AI for transparency and diversity. To date, the program has been downloaded more than one million times on Hugging Face.

• Trustworthy AI without trusted data

Federated learning seeks to solve the security issues associated with centralized data, which feeds large language models. However, it remains vulnerable to erroneous or malicious data. Developed by the Distributed Computing Laboratory, the ByzFL library strengthens the robustness of these systems by filtering out extreme values, which is essential for deploying truly secure AI in critical applications.

• Swiss research is back on the European stage

After years of uncertainty, Switzerland has regained its status as an associated third country with European programs retroactively to 1 January 2025. Swiss academic institutions can once again participate fully in Horizon Europe, Euratom and Digital Europe. In 2025, EPFL-based researchers were chosen for 33 Horizon Europe grants, seven ERC Starting Grants, as well as seven ERC Advanced Grants under the 2024 call for projects.

• A robot that adapts to the terrain

A bioinspired robot developed in the CREATE Lab can change shape to alter its own physical properties in response to its environment, resulting in a robust and efficient autonomous vehicle as well as a fresh approach to robotic locomotion.

• Unraveling the brain's hidden motor modules

Scientists from the EPFL's Brain Mind Institute, the University of Cambridge and Kumamoto University, have identified previously unknown neural modules in the brain that control movement and adapt during skill learning. Their findings challenge long-held ideas about how the brain organizes movement.

• Soft brainstem implant delivers high-resolution hearing

The Laboratory for Soft Bioelectronic Interfaces has developed a flexible auditory brainstem implant that closely conforms to the curved surface of the brainstem. The technology has been successfully demonstrated high-resolution "prosthetic hearing" in macaques.

• An injectable hydrogel for local bone densification

By combining injections of a novel hydrogel with systemic osteoporosis drugs in rats, a team from the Laboratory of Biomechanical Orthopedics has achieve rapid local increases in bone density. The results offer hope for future fracture prevention therapies in osteoporosis patients. The injectable hydrogel has been developped by the EPFL start-up flowbone.

• More selective and less costly carbon capture

A team at the Gaznat Chair in Advanced Separations has developed a scalable technique to create porous graphene membranes that selectively filter CO₂ from gas mixtures. Their approach slashes production costs while improving membrane quality and performance, paving the way for real-world applications in carbon capture and beyond.

• Tire additives on our plates

Traces of the additives typically used in tire manufacturing have been detected in all of the most common types of fruits and vegetables eaten in Switzerland. That's the key finding of the Central Environmental Laboratory and the Swiss Federal Food Safety and Veterinary Office study. The scientists don't yet know the long-term implications of exposure to these substances for human health.

• AI monitors wildlife behavior in the Swiss Alps

EPFL scientists have collected and curated MammAlps, the first richly annotated, multi-view, multimodal wildlife behavior dataset in collaboration with the Swiss National Park. MammAlps is designed to train AI models for species and behavior recognition tasks, and ultimately to help researchers understand animal behavior better. This work could make conservation efforts faster, cheaper, and smarter.