Florian Tramèr

SwitzerlandscientistMale6.5Global Dominance: 0.00%

Florian Tramèr is a professor at ETH Zurich, specializing in the security of artificial intelligence systems. His research focuses on the vulnerabilities of AI agents, particularly the risks associated with 'prompt injection' attacks, where malicious commands are hidden within data inputs. Tramèr argues that the mixing of commands and data in AI systems poses significant security challenges, highlighting the need for stricter separation to prevent unauthorized actions. He is also a scientific advisor for Invariant Labs, a startup dedicated to addressing AI security issues, and has discussed the implications of AI manipulation within academic contexts, recently gaining attention for his insights into hidden messages in research papers aimed at influencing AI evaluations.