Robust Machine Learning with Tobias Lorenz

Modern AI models are powerful, but also vulnerable. They can be manipulated not only during their use but even during training by feeding them tampered data. This can lead to dangerous misjudgments. In this episode of CISPA TL;DR you’ll learn how the new FullCert tool can mathematically prove that a model is secured against such attacks—from training all the way to deployment. CISPA researcher Tobias Lorenz explains exactly how it works and why this is crucial for our digital security. Tune in to find out how his work is helping to make AI more trustworthy. The episode is in German and available on all major podcast platforms.