How to hack an AI and why you should care
'Good' hackers managed to outsmart AI models in order to improve security. Hacked AI could tell you how to build a bomb or steal your private data. How do you get an AI to go rogue? And why is the problem so hard to fix?