Forget those "safety" myths! As AI goes full throttle into automation, our privacy and security stakes are through the roof. Expert Katharine Jarmul spills the tea on why those fancy AI guardrails are surprisingly flimsy and how models are totally memorizing your sensitive data. Engineers, don't just kick back and expect model providers to magically fix your privacy woes!
You've got to step up! Things like simple renaming tricks can bypass filters, and red-teaming once isn't cutting it. Instead of waiting for the next AI update or blindly trusting providers, it's time to build a solid, interdisciplinary security radar, consider local LLMs, and seriously diversify where your AI comes from to really catch those nasty incidents before they blow up.
Watch on YouTube
Top comments (0)