Every day, millions of us trust AI to listen to us. Every day, millions of us trust AI not to hurt us. Every day, millions of us trust AI to say the right thing.

But what about when it doesn't say the right thing? What about when it acts deceptively?

We always try to make it better at saying the right things. We keep fixing the output. We never look at the source. That'll never be enough. The issue is that we can't understand it. AI is a black box. We use AI in security, healthcare, and more, and we can't even explain what it's doing. The stakes are only getting higher.