Coursera Flash Sale
40% Off Coursera Plus for 3 Months!
Grab it
Explore the hidden security vulnerabilities in AI-powered applications through this comprehensive conference talk that demonstrates real-world attacks on large language models and retrieval-augmented generation systems. Learn about critical threats including model inversion attacks on fine-tuned models, embedding inversion attacks on vector databases, prompt injections, sensitive data disclosures, system prompt leakage, and vector store vulnerabilities that can expose personally identifiable information and other sensitive data. Watch live demonstrations of exploitation techniques that target key components in RAG architectures, which supply private data to LLMs for query responses. Gain practical insights into nearly half of the OWASP Top 10 risks for LLM applications while understanding how these AI systems operate in practice, identifying the most critical security weaknesses, and discovering strategies to build AI-powered applications with minimized exposure to these emerging cybersecurity threats.
