Natasha Tiku

Yeah, so we just wanted to, I guess, like, clear the air a little about, like, what is a jailbreak and...

Because Anthropic was throwing around the term, David Sachs was throwing around the term, but folks who had actually seen the reports that were shared with government officials, like in one case, Katie Masouris from Luda Security, she said the report she saw just said, fix the code.

And it fixed it, which it wasn't supposed to do, but this is not like a super sophisticated operation here, like.

you know, these things happen.

It's not possible to 100% secure a generative, especially a generative AI model.

potentially a bunch of questions around cybersecurity, but there's infinite ways you can ask these questions.

So like the data that they use for safety training is never going to be as big as like the data sets that are used to train the model.

So you're kind of depending on, you know, folks in the Bay Area to come up with all the possible scenarios and try to protect against that.

I don't know.

Yeah, we're living in their world, I think.

But at the same time, though.

Agreed.

It's a total mess.

But I would say a lot of those concerns are back in the limelight.

But I will say I was not expecting to have to still be talking so much about AI safety concerns in, what is it, June 2026.

As an AI scientist, entrepreneur, and investor, I know what it takes to build AI that works for everyone.

Every week, I sit down with the pioneers shaping our future, and we take you behind the scenes of the AI that's transforming our lives.

Find Pioneers of AI wherever you tune in.

Oh, Theo Baker.

Yes, it did.