Explore emerging vulnerabilities in large-scale Natural Language Processing (NLP) models in this hour-long conference talk presented by Eric Wallace from the University of California, Berkeley. Delve into the potential security risks, privacy concerns, and insights that arise from the increasing scale of modern machine learning and NLP models. Examine how adversaries can exploit these vulnerabilities to extract private training data, steal model weights, and poison training sets, even with limited black-box access. Gain valuable perspectives on the impact of model scaling and its implications for the field. Learn from Wallace, a PhD student supported by the Apple Fellowship in AI/ML, as he shares his research on making large language models more robust, trustworthy, secure, and private.