Heads up! As AI goes full throttle into automation, our privacy and security are on the chopping block. Expert Katharine Jarmul spills the tea on why those fancy AI "guardrails" are basically just wishful thinking and easily bypassed. Engineers, don't expect model providers to magically fix privacy for you; these models are like digital sponges, memorizing your sensitive data thanks to some architectural quirks.
Forget one-and-done security checks or hoping the next AI update will save the day. Jarmul stresses the need for a culture where risks are caught early, security is an ongoing iterative process (think threat modeling!), and maybe even diversifying your tech stack with local LLMs instead of relying on a single provider. It's time to get real about AI safety!
Watch on YouTube
Top comments (0)