Imagine a world where AI is sharper than even the most skilled hackers, catching malicious intent in milliseconds and thwarting cyber threats with the precision of a laser. Yet, somehow, these digital masterminds are still baffled by a well-timed cat meme or a furry friend caught in an image classification glitch. In this talk, we’ll take a lighthearted journey through the highs and lows of securing AI, exploring the surprising vulnerabilities that lurk in even the most advanced systems.
You’ll see demos where we challenge AI models with innocent (and hilarious) triggers, shedding light on what makes AI vulnerable to odd, even silly scenarios. We’ll also look at practical implications—how these weaknesses could actually be exploited in real-world security threats and what we’re doing to stop it.
Through live examples and interactive elements, we’ll explore how the quirks in machine learning can open doors for exploitation. This will lead us into a deeper understanding of the essential balance between building powerful AI defenses and safeguarding them from unexpected hiccups. We’ll wrap up with a few tips for those designing AI systems to ensure they can distinguish between a hacker’s attack.
Join us to laugh, learn, and leave with fresh ideas on making AI secure—and a little less confusion by cats.