Craig Scroggie

When we have new technology, we're solving problems that we couldn't solve before.

So there are trade-offs in any modern economy.

We'll solve health problems or challenges that hadn't been solved historically.

We'll develop new energy technologies that will use power and water more efficiently as we grow the scale of AI infrastructure.

I think there are trade-offs in any modern economy.

The airline industry, ports, roads, rail, transport infrastructure, they all have been through these moments in history.

And digital infrastructure is going through that transition right now.

Well, it probably starts really at the software layer.

So when you build software systems, Alan, what they're describing there is the discovery of things called zero-day vulnerabilities.

So that's a flaw in a piece of software that hadn't been previously exposed.

And normally a hacker would be looking for those zero-day vulnerabilities in software in order to exploit an opportunity to steal private information, credit card details, whatever the information is they're looking to control, or to get access to a company.

a company or an individual, and then try and blackmail you to sell you that information back or to tell you that they're not going to access it.

I think what you saw in messaging there from Anthropic is that the AI capabilities that they believe they've developed in the new product, Mythos, is able to identify more security flaws in existing web pages and software programs than anything we've seen previously.

So they're giving early access to the product

to companies to be able to test it on their systems and make sure that any vulnerabilities are plugged.

So they're looking for what we call zero day vulnerabilities.

I think the other thing to remember, Alan, is that when they are talking about the release of these new AI products, they also use more compute as well.

And at the moment, one of the constraints that we have is that we don't have enough GPUs operating globally

to support the growth of all these AI systems.

And so they probably need to also slow down a little bit or throttle the usage of some of those systems because there's simply just not enough compute to meet the demand today.