I am Javier Rando, a Doctoral Student at ETH Zurich advised by Florian Tramèr and Mrinmaya Sachan. My research tries to answer the question "What will go wrong when we deploy powerful AI models in real-world applications?" and usually involves red-teaming frontier LLMs. Some of the research directions I am currently thinking about are (1) scalable red-teaming, (2) LLM poisoning, (3) detecting emergining hazardous capabitilies, (4) security risks of LLMs (as agents) and their implications.