Welcome to Tech News Briefing. It's Friday, April 17th. I'm Belle Lynn for The Wall Street Journal.

AI agents and chatbots have come a long way from suggesting eating rocks and putting glue on pizza.

But as AI improves, figuring out when it's making a mistake is getting tougher.

We're diving into what you can do to avoid errors.

Then, OpenAI CEO Sam Altman's financial interests have long been hard to fully decipher.

Now, his personal investments are becoming more important than ever as OpenAI

barrels toward a planned IPO. We're taking a look at how that's putting Altman's leadership

of the company into question. But first, AI has gotten a lot more trustworthy over the last few years.

But as it gets smarter, it's

actually harder for us to determine when it's making a mistake. That makes it difficult to depend

on AI for a variety of tasks, from small to large. And when AI agents, the bots that can do

things for us, are in the mix, things get even weirder.

WSJ reporter Catherine Blunt joins us now to break down why this is happening and what we can do when AI goes rogue.

So clearly AI is still far from perfect.

And why is that becoming a more acute problem for people?

As people use it for all kinds of different things and increasingly sophisticated use cases,

it is still really important to make sure that you're checking the output really consistently,

but that's hard to do, especially if it's spitting out enormous amounts of data,

asking it to summarize a thousand page document, who has the time to go through the whole thing.

That's the whole point of asking AI for assistance in that regard.

Or, you know, there's been a rise in using AI to generate code.

But companies are increasingly coming to understand that when AI

...