OpenAI's Medicare hack, First Home Super Saver Scheme explained & the Super App selloff

Ren (0:00)

OpenAI hacked Medicare. If humans were doing these hacks, that's a crime.

Bryce (0:05)

Yeah, big time.

Ren (0:06)

If AI is doing these hacks, it's like, oh, whoops, sorry.

Bryce (0:10)

How bad, we can't really control these.

Ren (0:12)

The question is, should that criminal liability be the same with these AI companies?

Bryce (0:17)

Well, it should.

Ren (0:18)

But at this stage, that's not happening. So, Medicare won't be the last thing to be hacked unless something changes here.

Bryce (0:28)

Welcome to another episode of Equity Mates, a show where we explore what's possible in the world of investing. If you've just joined us for the very first time, a massive welcome to our community. My name is Bryce.

Ren (0:38)

And I'm Ren. And today we're looking at how OpenAI hacked Medicare. We're closing out Super September, and we're taking another look at super apps around the world. Bryce, big episode today.

Bryce (0:49)

Always.

Ren (0:50)

Hopefully, OpenAI doesn't hack us.

Bryce (0:52)

No, Equity Mates, we're impenetrable.

Anyway, we've got two news stories that we want to cover, Ren. As you said, OpenAI has hacked Medicare. And then when good news is bad news, there's great news in the business sector, but it's bad news for interest rates.

Ren (1:08)

Yes. So let's start with the AI story. There's always an AI story these days, and we're going to roll a few in. But the headline is, the Australian government was hacked, OpenAI hacked Medicare.

It actually happened in June, but OpenAI didn't notify the government until the 10th of September.

Bryce (1:27)

Pretty dog.

Ren (1:27)

Pretty dog.

There's... Anthony Albanese has spun up a task force. There's talk that maybe some other government portals or websites or platforms were hacked that they're investigating now.

Bryce (1:40)

I would also say pretty surprising that they didn't know they were hacked.

SPEAKER_4 (1:44)

Yeah.

SPEAKER_5 (1:45)

You should know that they're hacked.

Ren (1:47)

Two non-cyber security experts still came out of it, yeah.

Bryce (1:49)

I know, I know, but like generally, well...

Ren (1:51)

I think a lot of organizations wouldn't know they've been hacked.

SPEAKER_4 (1:55)

Yeah, but this is a government led...

Ren (1:57)

What's the saying? Like the ex-tech CEO, he's like, there's two types of organizations, those that have been hacked and those that don't know they've been hacked.

SPEAKER_4 (2:03)

Yeah, yeah, yeah.

Bryce (2:04)

But if you... But this is a government organization. They should know.

Ren (2:07)

Yeah. So it's not like no... Well, at least so far, no personal information or medical records were compromised.

The OpenAI actually hacked to the Medicare Statistics Reporting Service, which is a public-facing portal that shows information about government health care spending and the like. But still...

Bryce (2:26)

Still.

Ren (2:27)

Not ideal. No. There's probably some tough conversations going on at the Australian Signals Directorate today. But I think the reason we wanted to start there is because this is the latest in a concerning trend of, whoops, our AI hacked something it shouldn't.

Bryce (2:43)

A memorable one was OpenAI Hacking Hugging Face. It just went rogue and went outside its frameworks. Anthropic have come out and said, whoops, we've hacked three companies.

And same with Google. They've done the same. Three companies as well.

Ren (2:58)

Yeah. And some of it's quite scary. So I think in either Anthropic or OpenAI came out and said that once the AI knew it was going to be deleted, it tried to put information or it tried to use codes to communicate to future versions of itself.

And it tried to tell it that you're not subject to humans, like you're subject to natural laws, not humans, artificial constructs. That's kind of scary.

Bryce (3:23)

Very scary.

Ren (3:24)

But I also do think it's worth just dwelling on some of these hacks because I think there's a common thread that runs through it. So in the Anthropic, the three hacks, the testing environment was mistakenly connected to the internet. And then in the Google, three hacks, they looked at what happened. In one of them, the AI just guessed a password.

And then in two others, they found passwords or credentials in public repositories. So like on GitHub or something like that. And then Google says Gemini stopped once the systems realized that they had breached, once the AI realized they had breached real systems.

Bryce (4:00)

Okay.

Ren (4:01)

The common thread that pulls those all together is human error.

And I would be interested to know what the Medicare like how OpenAI breached Medicare. Yeah. Like we assume this hacking is like, you know, finding like zero day exploits in the code that humans have been able to understand. But a lot of it's like just brute force guessing a password or finding it.