Let's talk about it- AI for Security, Security for AI, or Security from AI?

Hi! Everyone is talking about AI, security, AI security, AI for security, and security from AI. Surprisingly, these are all different conversations, and each is equally fascinating in its own right.

In this talk, we’ll untangle (or at least take our shot) the surprisingly messy world of AI for Security and Security for AI, without hiding behind buzzwords or acronyms to sound smarter than we are. We’ll explore how researchers and practitioners are using AI to discover vulnerabilities, generate exploits, and fuzz applications in ways that are far more intelligent than traditional approaches. Then we’ll flip the script and ask a more uncomfortable question: what happens when the AI becomes the thing that needs protecting?

From prompt injection and jailbreaks to model stealing, data poisoning, and AI guardrails, we’ll look at the attacks that keep AI researchers awake at night. Along the way, we’ll peek into emerging research areas like LLM-based fuzzing, autonomous red teaming, and Verifiable AI- where, instead of simply trusting an AI’s answer, we ask it to show its work… or even prove it’s right. I am currently working my research on the topic of verifiable AI- and if time permits would love to present my observations on the same! I am a keen researcher, and have 2 patents, and 4 scopus indexed papers to my name.

Whether you’re a security enthusiast, an AI practitioner, or simply someone who’s wondered why people keep trying to convince chatbots they’re their grandmother, or generating ridiculous images of themselves- or someone just curious about what even is the hype? I hope you’ll leave with a clearer picture of one of the biggest shifts happening in cybersecurity today: the future isn’t just about using AI to defend our systems- it’s about defending AI, too.

Join me on our quest to make sense of a rapidly evolving field…