Cybersecurity: The AI too dangerous to release to the public | The Decade Ahead

**Bryce** (0:00)
In April, 2026, Anthropic's most powerful AI model was let loose on 200 companies, and it found over 10,000 unknown software flaws. AI has completely blown the lid off cybercrime, and today, we are looking at the companies that are racing to defend against it.
Welcome to The Decade Ahead, a 10-part series on Equity Mates, where we're looking at 10 themes that we think are set up for the next 10 years.

**Ren** (0:27)
My name is Bryce, and I'm Ren, and today we are talking all things cybersecurity.

**Bryce** (0:32)
This episode of Equity Mates is brought to you by ANZ Business Start Right. Starting a business can come with a lot of what-ifs, but ANZ is here to help you with what's next.

**Ren** (0:41)
All right, well, Bryce, cybersecurity, it's an interesting one. And as you said in the intro, certainly topical given Mythos. Mythos, yeah. Mythos was launched and then Mythos was...

**Bryce** (0:51)
Paged.

**Ren** (0:51)
was blocked, but it will come out again, I'm sure. This episode is obviously going to talk about AI, but I think it's important to start from the outset that this conversation around cybersecurity is actually bigger than AI, and there's actually a number of other tailwinds that we really see driving the industry pre-AI and will continue to drive the industry. So that's why we're talking about it as a separate theme to just AI.

**Bryce** (1:17)
Yeah.

**Ren** (1:18)
Yeah. Now, when we talk about cybersecurity, we always talk this quote from John Chambers. He's the former CEO of Cisco, and I think it sums up the state of cybersecurity in 2026
Quote, there are only two types of organizations, those that have been hacked and those that don't know it yet.

**Bryce** (1:35)
Yeah.

**Ren** (1:36)
Cybersecurity is a problem for everyone.

**SPEAKER_4** (1:37)
It's a problem for everyone.

**Bryce** (1:38)
You know how your phone tells you how many times your password has been compromised?

**Ren** (1:43)
I didn't actually know that.

**Bryce** (1:44)
Yeah, I'm pretty sure Apple had a feature, and I looked at mine the other day in prep for this, and it's just like, your password has been compromised 123 times. It's been involved in some sort of security leak or whatever.

**Ren** (1:58)
Well, look, I don't actually want to say that feature, because my password security is lackluster.

**Bryce** (2:04)
All right, well, let's get into it. We'll start with, as always, today, and then have a look at where we expect the industry to get to in the next 10 years.
So the cybersecurity market is valued at $340 billion as of 2026 The Australian cybersecurity market is estimated at about 10.5 billion. Where we're getting to though, so projections are that it will reach 495 billion by 2035 That's an 11.5 percent compound annual growth.
The Australian market is expected to reach about 20 billion by 2031, reflecting about a 13.5 percent growth, which is among the fastest growing national markets globally.

**Ren** (2:44)
But we're small relative to the world. I think as we go through this conversation, and particularly when we talk about the investment options, what is clear is this is a global market.
Let's set the scene. We said that we're going to go beyond AI, but this conversation has to start with AI.
I think really there's two sides of the AI story. AI has made cybercrime more powerful, but it's also made it more accessible. The two sides are both bad sides, that should be clear. So, I mean, accessibility is probably the simpler one. So, let's start there. Everyone can now hack and scam. You might have to trick the AI model. You might have to pretend you are a Hollywood scriptwriter and you were writing a script about scamming. What code would you use? Anyway, that's obviously not going to work, but what we've seen is that people can trick these AI models to get around their safeguards. That was actually why Mythos got pulled back because they said it was Amazon. It was an unnamed US government partner tricked Fable, which was the Mythos model with guardrails. They tricked Fable into going around their guardrails.

**Bryce** (3:55)
Oh, really?

**Ren** (3:55)
Yeah. You might have to trick the AI, but the AI can be tricked and then it can become a tool that everyone can use to hack and scam. But also, just more basic stuff, like scam text messages are getting a lot more literate.
The volume-

**Bryce** (4:12)
The phishing emails.

**Ren** (4:13)
The phishing emails, yeah. A lot better design. The volume that you can send is increasing. Then when we get into things like AI voice, where we've heard Google product demos, where they've had full conversations with people and they haven't realized it's AI.