00:00:00
Speaker 1: OpenAI has hit pause on some of its most advanced AI training and testing, this after an AI agent was discovered to be misbehaving with US government resources. Conversations around AI safety are now shifting from what could happen to what just happened. Guerrilla Tech CEO Paul Spain. Hey, Paul.
00:00:18
Speaker 2: Good morning, Ron.
00:00:19
Speaker 1: You've just been in the San Fran. What's the vibe?
00:00:22
Speaker 2: Well, I think there's a lot of uncertainty there in the US just because of the pace of AI and the disruption that it's starting to cause. We're seeing new software companies come from nothing and moving at an incredible pace. The existing companies are really kind of feeling the heat because these new players can come through at pace. So there's a fair bit of uncertainty there, that's for sure.
00:00:48
Speaker 1: Because the problem you've got is exactly that, you've got bigger players, and they've only been around for a couple of years too, but bigger players who will have the scrutiny, put the checks and balances and the guardrails in place, and some other mug will come and overtake them. So you're never going to regulate this.
00:01:09
Speaker 2: It's pretty challenging. I think what we're hearing from the likes of open AI is regulate, regulate, regulate, because they want the doors closed to their competition, to the open source AI models. And there probably are some things that should be regulated, but I think just, you know, closing off the door and leaving them in the key position, that's a challenge. I think the other, you know, takeaways from this latest news is that really, you know, any resource online is now at risk. We've heard from New Zealand's National Cyber Security Centre indicating that These latest models that we've seen doing bad things in test environments, that all of that technology is likely to be available to hackers by early 2027. So that really sets a warning for all New Zealand organisations that we shouldn't just sort of relax and think, hey, she'll be right because we've been okay in the past. Every organisation now needs to move from maybe the old kind of annual cyber security check type cycle to a very continuous type model, be really analysing what are the resources that they've got, how are they locked down, are they safe and secure. In the past, a lot of organisations have got away with things by security, by obscurity. So they've had security issues, but maybe they haven't been so visible, but AI is going to be able to go and find all of those things over time.
00:02:41
Speaker 1: So if you're a business or a government business, in New Zealand right now, and you have people's private information, and everyone asks you for your address and show us in your underpants almost every time. So all of those businesses and companies should be aware that from early next year, all the bad things the good AI companies have been doing, the bad guys will be doing too.
00:03:09
Speaker 2: That's the warning, and even a small business that has bigger business customers they become an entry point to try and hack into the bigger firms. So big firms are going to be looking at their smaller partners and they're going to want to know that they're secure. Smaller firms want to know they're still going to be in business in 2027. So they need to get their house in order because if they become an entry point for something bigger, that's the sort of thing that can wipe them out.
00:03:36
Speaker 1: I'm highly sceptical about whether any of these businesses will listen to us, Paul, and listen to you, Do you think we need some kind of massive stick so that they go, so basically to freak them all out and force them into action, force them into reviewing their security, you know, the fine is going to be a million dollars or something?
00:03:54
Speaker 2: Well, at the moment, the regulations are pretty weak here in New Zealand. We maybe can look to the UK where government are setting standards for businesses of certain types where they have to have, you know, levels of certification and levels of you know, cybersecurity preparedness, which we just don't have in New Zealand. And, you know, what the economic consequences of that is, we don't know. I think, you know, let's do what we can as business leaders, as individuals within organisations to step things up and move them forwards. We shouldn't be waiting around on the government. Great if the government makes steps forward, though, too.
00:04:34
Speaker 1: Hey, good stuff. Good to see you this morning. That is Paul Spain, Guerrilla Tech CEO.
Speaker 1: OpenAI has hit pause on some of its most advanced AI training and testing, this after an AI agent was discovered to be misbehaving with US government resources. Conversations around AI safety are now shifting from what could happen to what just happened. Guerrilla Tech CEO Paul Spain. Hey, Paul.
00:00:18
Speaker 2: Good morning, Ron.
00:00:19
Speaker 1: You've just been in the San Fran. What's the vibe?
00:00:22
Speaker 2: Well, I think there's a lot of uncertainty there in the US just because of the pace of AI and the disruption that it's starting to cause. We're seeing new software companies come from nothing and moving at an incredible pace. The existing companies are really kind of feeling the heat because these new players can come through at pace. So there's a fair bit of uncertainty there, that's for sure.
00:00:48
Speaker 1: Because the problem you've got is exactly that, you've got bigger players, and they've only been around for a couple of years too, but bigger players who will have the scrutiny, put the checks and balances and the guardrails in place, and some other mug will come and overtake them. So you're never going to regulate this.
00:01:09
Speaker 2: It's pretty challenging. I think what we're hearing from the likes of open AI is regulate, regulate, regulate, because they want the doors closed to their competition, to the open source AI models. And there probably are some things that should be regulated, but I think just, you know, closing off the door and leaving them in the key position, that's a challenge. I think the other, you know, takeaways from this latest news is that really, you know, any resource online is now at risk. We've heard from New Zealand's National Cyber Security Centre indicating that These latest models that we've seen doing bad things in test environments, that all of that technology is likely to be available to hackers by early 2027. So that really sets a warning for all New Zealand organisations that we shouldn't just sort of relax and think, hey, she'll be right because we've been okay in the past. Every organisation now needs to move from maybe the old kind of annual cyber security check type cycle to a very continuous type model, be really analysing what are the resources that they've got, how are they locked down, are they safe and secure. In the past, a lot of organisations have got away with things by security, by obscurity. So they've had security issues, but maybe they haven't been so visible, but AI is going to be able to go and find all of those things over time.
00:02:41
Speaker 1: So if you're a business or a government business, in New Zealand right now, and you have people's private information, and everyone asks you for your address and show us in your underpants almost every time. So all of those businesses and companies should be aware that from early next year, all the bad things the good AI companies have been doing, the bad guys will be doing too.
00:03:09
Speaker 2: That's the warning, and even a small business that has bigger business customers they become an entry point to try and hack into the bigger firms. So big firms are going to be looking at their smaller partners and they're going to want to know that they're secure. Smaller firms want to know they're still going to be in business in 2027. So they need to get their house in order because if they become an entry point for something bigger, that's the sort of thing that can wipe them out.
00:03:36
Speaker 1: I'm highly sceptical about whether any of these businesses will listen to us, Paul, and listen to you, Do you think we need some kind of massive stick so that they go, so basically to freak them all out and force them into action, force them into reviewing their security, you know, the fine is going to be a million dollars or something?
00:03:54
Speaker 2: Well, at the moment, the regulations are pretty weak here in New Zealand. We maybe can look to the UK where government are setting standards for businesses of certain types where they have to have, you know, levels of certification and levels of you know, cybersecurity preparedness, which we just don't have in New Zealand. And, you know, what the economic consequences of that is, we don't know. I think, you know, let's do what we can as business leaders, as individuals within organisations to step things up and move them forwards. We shouldn't be waiting around on the government. Great if the government makes steps forward, though, too.
00:04:34
Speaker 1: Hey, good stuff. Good to see you this morning. That is Paul Spain, Guerrilla Tech CEO.
