AI's dive into full automation means our privacy and security are totally on the line! Expert Katharine Jarmul spills the tea on why current AI "guardrails" are easier to bypass than you think, and how models are basically data hoarders, memorizing all your sensitive stuff. So, stop expecting model providers to magically solve these privacy puzzles for you.
Forget waiting for the "next version" to magically fix everything. Instead, get proactive! We're talking about understanding why models memorize, embracing iterative security with things like threat modeling, and even considering local LLMs and diverse providers to keep your digital fort genuinely safe. Remember, speed and performance don't equal bulletproof security!
Watch on YouTube
Top comments (0)