Transcript
0:00 As AI becomes more powerful and pervasive in organizations and people's day-to-day workflows, security risks can grow to be more and more dangerous, as well as becoming more critical to mitigate. Understanding these risks and how to defend against them is not optional. I'm Neil Daswani, co-academic director of the Stanford Online Advanced Cybersecurity program. Together, we'll dive into the growing field of AI security and explore potential risks in the design, build, and maintenance of AI systems in our new AI Security course.
0:31 By the end of this course, you'll gain an understanding of how to make AI systems more secure and resilient in the face of modern threats. So now let's get into the specifics-- You'll learn how to evaluate vulnerabilities across the AI lifecycle, from training data poisoning to prompt injection, jailbreaks, hallucinations, adversarial examples, and other inference time threats. In this segment, we're going to talk about the different system architectures that are being used. We'll cover the architectures of modern AI systems, including multi-agent setups, and examine high-profile breaches involving large language models and deepfakes.
1:07 The big question that we have is, how much further can you push it? Guided by world-renowned faculty and industry experts from leading companies, we equip you with the knowledge and skills needed to safeguard your organization. I think everybody envisions this world where the machines are just going to write all the code. But I think what we're missing are all the things we're going to build around that. Join us in mastering the critical skills needed to fortify AI systems and empower their resilience against evolving threats.