Anthropic’s AI model Mythos reportedly possesses hacking capabilities far beyond those of most of the world’s top hackers. Does this represent an opportunity for global companies to strengthen vulnerable operating systems – or is it a powerful new tool that could further enable cybercriminals? Sustainable Investment Specialist Grace O’Hanlan discusses this latest AI development with Ryan Bridge, exploring why it has triggered emergency discussions at the highest levels of governments worldwide.

Click here to download the MP3 file or listen to the podcast on your favourite platform:

Bridge talks Business: 21 April 2026
Episode Transcript


Ryan Bridge
Kia ora and welcome to episode 72 of Bridge talks Business with Milford. While investor focus in recent weeks is rightly centred on the physical warfare unfolding in the Middle East, there have been some equally significant developments on a less visible front, cyber warfare. Artificial intelligence firm Anthropic has announced its new model called Mythos. It is strikingly capable at computer security tasks, and it has found vulnerabilities in every major operating system and web browser in the world. Grace O’Hanlon from Milford is here to break it down for us in our feature interview today. First, here’s your top five business bits.

    1. Global share markets performed strongly last week, rising on optimism of a US-Iran deal. Despite that optimism, the headlines continue to give mixed messages. There’s clearly a divide here between the more diplomatic government figures in Iran and the more extreme revolutionary guard. This is making life pretty difficult for the US.
    2. Partial inflation data Friday showed the first clear evidence of higher energy prices. Offsetting the strength in energy was more modest softness in other areas. Q1 data is out, but only includes a month of the conflict so far.
    3. Central bankers have been briefing over the past week. In general, speakers from the Bank of England, the ECB, the RBNZ, they all spoke in a more dovish tone. Markets had moved to price quite aggressively, tightening in those regions. RBNZ Governor Bremen, Bank of England Governor Bailey also noted that firms have flagged a lack of pricing power. Now that suggests more modest second round inflation could be on the cards.
    4. The UK Prime Minister, Sir Keir Starmer was again in the press for all the wrong reasons, with new information highlighting UK Ambassador to the US Peter Mandelson’s failed vetting ahead of his appointment. The PM claims this was something he was not aware of, but now faces a tough leadership challenge in the House of Commons. This is poor timing ahead of the local elections there this May, where Labour is tipped to lose a number of seats to reform.
    5. In the week ahead, markets will continue to watch US reporting season, of course. We also hear from Trump’s new Fed Chair nomination, Kevin Warsh. He’s going to present in front of the Senate Banking Committee. Now that approval is expected to be blocked until Trump withdraws the DOJ investigation into the current Chair, Jerome Powell.

For our feature interview today, what happens when you take cyber warfare and cyber crime and couple it with AI? What happens? What happens to a business, what happens to the security of information that a business holds? Same goes for governments. This is a real concern around the world right now and something we’re going to talk to Grace O’Hanlon about who’s a Sustainable Investment specialist at Milford. Just a reminder this segment is informational only and should not be considered financial advice. Grace, welcome back to the podcast.

Grace O’Hanlon
Thank you very much.

Ryan Bridge
Now, you’re here to talk to us about Mythos. We’ve already got through the pronunciation difficulties. It’s Mythos for those who are concerned. Tell us, what is it?

Grace O’Hanlon
So Mythos is Anthropic – so a big AI company – it’s their new, yet to be released, general purpose AI model. So when they were testing it, they realised that this model has vastly superior capabilities at finding and identifying and then exploiting software vulnerabilities, and being able to break into and ultimately hack systems. So it found vulnerabilities – reported by itself – but in every single major operating system, every major commonly used browser. So think Firefox, Linux, and all those sorts of things. It can autonomously, under testing conditions, break in there and start to wreak havoc with those systems. So Anthropic is saying that this model has the capability of all but the very, very best human hackers.

Ryan Bridge
So they’ve essentially created an AI hacker.

Grace O’Hanlon
Yes. That doesn’t eat, doesn’t sleep, doesn’t drink, and it can do all of these things by itself all at once. Basically, you’re giving the capability of the world’s best hacker, you’re giving that capability to individuals with much less skill, and so they can much faster on a much, much greater scale launch attacks simultaneously against some of the world’s most common operating systems.

Ryan Bridge
Sounds fabulous. Sounds like a great idea. Grace, why should we be concerned? Or how concerned should we be? Because it’s not public yet, is it, this product?

Grace O’Hanlon
No, that’s correct. And so Anthropic, who created the model and know most about it, they have reported that it is genuinely a fundamental concern for us right now in the digital world. So they have decided prudently not to release this model into the public yet. And what they have done is formed a specialised, trusted group called Project Glasswing, very tech. And they’ve invited all of the major tech firms, Microsoft, Amazon Web Services, etc. Big financial institutions, governments, cyber security specialists, all into this this group. And what they’re going to do is launch Mythos on these big operating systems, identify the weaknesses, and patch them before Anthropic is ever going to release it potentially to the public.

Ryan Bridge
Wow, is it not possible that someone else could get there first with this and it could be public anyway?

Grace O’Hanlon
And you’re absolutely right. And in the AI world, a day is a year in the old world. So of course, that’s true. And that is a huge risk. And it’s a risk that an AI company that is perhaps less ethical or a little bit less concerned about the impact, might race to publish something that does have pretty similar capability. So that is a huge risk. And I guess at the moment, we’re lucky that it’s Anthropic that has developed these capabilities, not maybe one of its peers that is a little bit less responsible and maybe a little bit more reckless.

Ryan Bridge
So obviously, there are massive risks. Is it possible that the risks could be worse than what the market is expecting could happen?

Grace O’Hanlon
That’s a great question. And it does depend on who you ask. And so I have been into some very deep, dark, very niche corners of the internet. And there is some commentary in there that say that this could be a Y2k moment for software. So in Y2k, you’re obviously too young to remember that, but I do. And we were ticking down to this point in time of ticking over to the year 2000. But in this case, the release of this really powerful model. And software engineers were racing to patch all these old, perhaps decades old, systems that may have these vulnerabilities or gaps. Or nobody knows what will happen as that clock ticks over. When that model gets released, who knows what might happen. So there is that risk out there. But how likely that is, yeah, that remains to be seen.

Ryan Bridge
Okay. Yeah. I remember the cars were going to suddenly stop when the clocks turned over at midnight. So what about other tests that are being done on the capabilities of Mythos?

Grace O’Hanlon
So that’s the thing. Basically, everything we know about Mythos to date comes from Anthropic themselves. So maybe this is a little bit overblown, perhaps. Anthropic – it is probably in their interest to exaggerate the capabilities of their mode,l because it is rumoured that they’re going to IPO perhaps in the next 12 months or so. And that’s great for valuation. It’s also great PR to say, “Oh, look how responsible and ethical we are. We’re foregoing revenue by withholding this really powerful model to safeguard humanity. We are the responsible owner. We should be a leader in this space and you should trust us.” So there is that. At the same time, you don’t want to be one of those companies that oversells the capability of your model. And then you found out later, and we’ve seen plenty of companies fall on their sword like that too. So I think to date, Anthropic has been pretty honest in its reviews of its prior models. So both their strengths and their weaknesses. So I think there and thereabouts, it can do pretty much what it says it can do.

Ryan Bridge
Grace, this is going to be huge – if true, and once we see the proof in the pudding I suppose – it has the potential to be a massive market disruptor, doesn’t it? I’m just thinking about all of those antivirus software firms, all of those companies that were built up through the ‘90s into the 2000s and we all install on our computers. What is the potential for this?

Grace O’Hanlon
Yeah. And again, I want to start, I might quote, the Canadian finance minister, he pointed at this and he said, “This is the unknown unknown.” So nobody really knows until we get there, which is true of many of the applications of AI to date. But what could happen? There’s a lot of legacy systems that are operating at the moment. And one bank may have a legacy system operating with a slightly new piece of software and the way that they communicate may have vulnerabilities in it. I think what we’re going to see, is perhaps AI begins to move back into that software development space at the very, very genesis before we’re just rolling out software products willy-nilly and trying to force them together. I think you’ll see AI being used by the biggest, most powerful companies in the world, software firms, finance firms, who’s interest it is to ensure that their digital world is locked up tight. They’re going to invest heavily in ensuring that their software is going to be highly defensible, highly robust, very trustworthy at the front end. So I do think there is going to be a lot of disruption in the near term. And there already is, of course. But I do think that in the longer term, AI can be used for good and it’s going to be used by the biggest, most powerful, most well-resourced companies.

Ryan Bridge
Grace, this is always informative when I talk to you and I always appreciate your input. However, a little bit scary today. Can I just say, a little bit scary? In saying all of that, there are hackers out there already, right? This stuff is already happening in the world.

Grace O’Hanlon
Yeah, you’re absolutely right. So with or without Mythos, whether this gets released to the public at some stage or not, there are excellent hackers out there that are already utilising AI really effectively. And that is only escalating both in pace and in sophistication. Just last week, an Iran-affiliated hacking group hacked into utility digital infrastructure in the United States and disrupted water and energy flows. In March, Striker, a huge medical devices company was hacked. And AI was utilised and it entered into the system, turned off every device associated with that system and it impacted surgical medical devices that are being used and in critical operations for people. So it is real and it’s happening already. It will continue to happen and it will continue to escalate, just powered by AI as it is. So Mythos, whilst very powerful, is just one more tool. And as you said, there’s going to be another Mythos. There’s going to be OpenAI’s version or perhaps a Russian or a Chinese version that we don’t even know about yet. So these risks are really real. And it is really important that companies continue to focus on their cyber defenses and continue to exercise really, really cautious due diligence.

Ryan Bridge
Wait, was that supposed to make me feel better about it? Grace, good to see you. Thanks so much for being on the podcast.

Grace O’Hanlon
Thank you very much.

Ryan Bridge
That was Grace O’Hanlon, Sustainable Investment Specialist at Milford talking to us about Mythos and also cyber security more generally Just a reminder you can like, follow and subscribe this wherever you like to listen to your podcasts. Until next week, don’t forget to invest in yourselves.

Missed previous episode? Don’t worry! Click here to catch up now.