Technology & Security Podcast with Dr Miah Hammond-Errey

Welcome to Technology and Security. TS is a podcast exploring the intersections of emerging technologies and national security. I'm your host, Dr. Mia Hammonderry. I'm the founding CEO of Strat Futures Pty Ltd and an adjunct associate professor at Deakin University.

Dr Miah Hammond-Errey (00:00)
My guest today is Nicholas Me I. Thanks so much for joining me, Nico.

Nicolas Miailhe (00:04)
Thanks from Vemia.

Dr Miah Hammond-Errey (00:05)
Nicholas Mehite

Works at the intersection of frontier AI security, safety, and international AI governance. He co-founded AI Safety Connect, a convening platform which aims to improve security, safety, and governance of advanced AI systems He co founded and leads Prism Eval, a Paris based AI safety startup specializing in evaluations and interpretability for advanced AI models and systems.

Previously, he founded The Future Society, an AI think tank in 2014 at Harvard Kennedy School of Government. He led that as CEO until 2023. And prior to that, he spent 15 years in defence and defence industry working on industrial cooperation and technology transfer.

He has taught postgraduate courses, held multiple professorial visiting roles. He is an AI expert for OECD and UNESCO's AI ethics group. Thanks so much for joining me on the Technology and Security Podcast, Nico.

Nicolas Miailhe (00:59)
Thanks for doing it.

Dr Miah Hammond-Errey (01:00)
We're coming to you today from the lands of the Gadigal people. We pay our respects to elders, past, present, and emerging, and acknowledge their continuing connection to land, sea, and community.

we're recording in June 2026, and in the last few days, Canada has released AI for all, an AI strategy. I won't get you to dissect it, Nico, but I know you've spoken a lot about relationships in in a range of different countries, and obviously a a large part of your work is about convening. So I want to go straight to a segment on alliances. What are the key alliances in AI and governance and what do we need to be focused on?

Nicolas Miailhe (01:37)
there is clearly the emergence ⁓ of a G2, where there are two AI superpowers, ⁓ US and China, it's probably a bit exaggerated, ⁓ because I think that there is space for more than that.

But as a matter of fact, the critical size of the market and the ⁓ investments deployed creates a G2. Okay, and that's a that's a that's a moment where even the the president of the US is really seeing this relationship in the field of economic development and trade and investment, including in the in the field of AI. And so a question is asked to those who are not part of this G2 who were

up to now part of other larger group G7 ⁓ which is also incorporating a number of values besides economic status such as the commitment to democratic liberal democratic modernization process and the G20 as a as a much wider group of emerging power. And so it's a question being asked to the old alliances over how they're going to have to reorganize in the age of AI. And that's why you've seen and how you've seen

From Canada itself at Davos emerging the notion of middle towers, as a concept around which a new and different form of coalition can forge. And but basically what these countries have in common is that by itself, the size of the population, the size of the market, the size of the investment that they can muster, and the way in which they are economically integrated.

Does not suffice to guarantee per se a trajectory where they can keep pace with the mastery of the tech stack that enables them to remain ⁓ in control of their future, to have leverage. This is usually formulated under the concept of sovereignty. However ⁓ complicated and blurred this concept is, also for good reason, because what does it mean to be sovereign in the age of AI?

Which emerges from the ground of digital platform, data, ⁓ networks, connectivity, all of these are already extremely integrated economically. And so the level of independence is huge. And so remaining sovereign in the age of digital, like it doesn't mean autarky won't work. And so trying to find the right ⁓ definition of what that means to be in control of your future, and defining the right kind of alliances in terms of economic interest.

political interest and geostrategic interest that point towards the future is really what the this this notion is trying to to point at. And so middle powers in the age of AI points to that. Now maybe calling out one particular ⁓ set of alliance of middle powers is obviously the European Union. Why? Because in and of itself it's already a collection of middle powers which have tried to organize themselves

In the age of the Cold War, in the balance between the US and the Soviet Union, and now finding itself projected in a world where it used to be at the center, it's not clearly at the periphery. In the age of AI, it's really the Asia-Pacific ⁓ economic region which is leading the rise of the AI value chain, And so that's just one example, but there are others, and Canada is clearly trying to

lead out of that and lead out of that out of a a newly discovered feeling of distrust and diffidence vis-a-vis who vis-a-vis the US. The way in which the US has decided to, under the current administration, to reorganize its alliance and its friendship with old friends and allies on its continent and beyond is stirring discomfort, strategic discomfort.

in Europe vis-a-vis the way in which ⁓ the war on Ukraine ⁓ and ⁓ dealing with the potential of an aggression, ⁓ a deepened aggression of Russia is handled, and beyond ⁓ trade and investment. So it's really let's also call it out. The rise of middle power is the rise is a reorganizing of the Western Alliance.

Dr Miah Hammond-Errey (05:22)
And it's so fascinating that this is coming at a time when, as you say, sovereignty is is really being tested and questioned.

You launched the AI Safety Connect at the Paris AI Action Summit in February 2025 with the conviction that the coordination gap on Frontier AI was closable. So eighteen months later and after convening many gov global forums, is that conviction weaker or stronger?

Nicolas Miailhe (05:47)
Much stronger, much stronger, but that has now taken an even more dramatic turn. That I would frame in the following words: failure is not an option. It's the picture in terms of how AI capabilities are gaining in strength, in abilities, ⁓ pointing towards artificial general intelligence as ⁓ as all of this is moving at at lightning speed to the point.

That ⁓ Entropic AI, ⁓ co-founder and now head of ⁓ Anthropic Institute, which has been created as a spin-off of the main company, to think about the immediate future as we are on the threshold of AGI, is talking about the need to slow down ⁓ and to to suspend some of the developments. I think this kind of alarm calls can be seen as a ⁓ fundraising scheme. I think they are.

mi because this company is also ⁓ on the verge of ⁓ IP tradition in competition with OpenAI. So these ⁓ communication signaling functions are also important to be ⁓ monitored, but I think it goes beyond it. I mean we have all seen in all in our in our day to day practice how things are getting better. And I think we have all been a bit astonished at the pace at which this technology, which again is difficult to secure ⁓ by design, is being agenticized across all economies.

At the center of the the information and knowledge society. So pointing to the the imperative of international coordination. Again, this revolution is not a local one, it's inherently a global one, which has been ⁓ captured in the in the narrative by this impression of a G2, which is extremely suffocating for all the others which are not part of that. The good all if you are not at the table on the menu, well, the table is growing smaller and smaller.

In terms of who's got actually something, who's got a chair at the table. And there are a lot of people watching the tables and passing notes to the people at the table. but I think it's when we think about this metaphor, G2, G20, G7. These are tables, these are moments which are ⁓ like ⁓ in terms of a vision represented as tables and and and and this matters. So the need for ⁓ the G2 to coordinate, those who are capable of operating at the frontier to

Exercise responsibility in how they're pushing the frontier, the way in which the others are exercising pressure for them to do that, and also the way in which others seek to bring to their economies and to their nation-building projects more of the leverage and agency over controlling what this fundamental revolution ⁓ means to them. ⁓ pluralism, diversity. Like we're talking about large language models, obviously mostly driven in for the West, that I know better in English.

Is that acceptable? Probably not. Is it acceptable that we are going to build our nations and our schools? ⁓ calling API from Entropic Open AI? wait, I mean this would be deluding ourselves. This would be this would be literally crazy in terms of lack of agency and dependency. So ⁓ developing the spectrum, developing ⁓ strategic optionalities to regain abilities to acquire more control over certain key capabilities from national security in the field of cybersecurity, for example.

To economic development is crucial. And all of that requires ⁓ international coordination and cooperation.

What type of externalities are we pushing on societies for our kids, for workers, for our ⁓ officers like survey, who needs to to protect from our borders to our communities? ⁓ so this notion of safety and security, which is key to again, it's not anti regulation. Either a system is liable and it goes through the critical value chains and it powers ⁓ cars and and air flies and it's used in in medicine, or it's not. So it's also a question of that, of economic development, it's a precondition.

For massification of the economic development and ⁓ and sovereignty, so which is the ability to have control and and oftentimes what we have seen so far is a tension between safety and sovereignty because of the desire to catch up at the frontier. And so a lot of what has happened is the sacrificing of safety and security because of the need to innovate fast to remain relevant at the frontier. And so dealing collectively, ⁓ including with this global systemic risk that we are causing,

Like a mythos model is capable of projecting systemic cyber risks on financial institutions that has consequences on the way in which ⁓ still most integrated ⁓ global governance system in the domain of finance is operating and is going to operate. And that's really the foundation of globalization. Finance remains the globalization of the remains the foundation of globalization. And so it's we at ASFTG have tried to

move the needle in how we reconcile what happens at the frontier, how do we provide more control ⁓ to ⁓ to nations which is legitimate vis-a-vis what is their cause otherwise otherwise they'll find ways around. They will steal. They will distill. And and so it's you know, because this technology in of itself is tough to secure, as I said, it's proliferative in nature. It's good because it's democratic in nature, if you use well, because it can circulate quite well.

We need and so that's why this tension is very important to reconcile and failure is not an option because it has now become strategic and geo-strategic in the context of an intense ⁓ competition between ⁓ and across powers, and at a time where ⁓ because of the way in which the US has decided to ⁓ shake up the tree, good old institutions of multilateral coordination are being put in question, are being reformed.

I would say never let a good crisis go waste, but that and and therefore reform these institutions. But that won't happen by chance. It will happen in a way as a function not only, not only, but as a function of the volume and the quality of our international coordination at different levels. Like obviously it's multi-stakeholder, it's not only policymakers. It's not international only, it's it's global. ⁓ but we also need, absolutely, because of its geographic nature, ⁓ policymakers at the center.

and it is to work from the it is to go from the working level to the top. Like we need political leadership. That's what Lashley part was about. And we need also capacity building at the working level and and and across. It's really a function of international coordination that has to be non-naive, that has to be humble, and that has to be quite intense. So to summarize, at a time where we need ever more international coordination, it's becoming more difficult because of the crisis of multilateralism.

Dr Miah Hammond-Errey (12:22)
I know you've spent ⁓ some time working on inclusive AI strategies for a number of different countries. When you look forward, say, two to five years, what are the countries that are not countries that we think of as being key in this debate which will actually be huge movers of the future?

Nicolas Miailhe (12:39)
It's a the positive deviance, in

Dr Miah Hammond-Errey (12:42)
not because they're necessarily gonna be at the frontier, but countries who have, large populations who are able to adopt new technologies quickly, and, you know, have large markets essentially that we don't think about necessarily, because we're still, bound in that old kind of power structure of globalization and those international relationships.

And one of the exciting things about technology, as you say, is the opportunity for reconsidering the way that we do things.

Nicolas Miailhe (13:10)
One one example Korea and one explanation as to why Korea is so interesting and and that drives or that that that that points in the direction of what are the things we should monitor and look for. Korea is a history of a fantastic industrialization and fantastically successful industrialization ⁓ policy and nation building project over the past forty years, even more. When I was a little kid growing up in Paris, Korea means what? Like toys in plastic? And now

It means a a fantastic position at the center of the ⁓ semiconductor value chain, coupled with a very impressive positioning at the level of terminals, from smartphones to networks, ⁓ to compute, at the application level, much less. Well, keep cap up also culturally, but it's a great point because culturally that too ⁓ in parallel of a of an enduring struggle for and towards democracy.

Dr Miah Hammond-Errey (13:54)
I was expecting you to say K pop.

Nicolas Miailhe (14:07)
That too exposed to a military existential threat. If you've been to Seul, you know the DMZ is what? At the shelling distance, when with 155 mm good old calibre 42 52 ⁓ Howitzers, like it's shellable. And so the distance, so we talk about missiles, but we're gonna also talk about more traditional artillery. So it's this distance is creates a strategic mindset of its own nature and a pressure, enduring pressure.

Of the military question in the way in which the nation is built and how democracy is organized as a collective ⁓ preference ⁓ selection model and optimization model. so I think that's a that that's a good example of a surprising because it's it's a relatively small economy that has ⁓ succeeded at ⁓ deploying itself and integrating itself economically ⁓ from Europe to the US and ⁓ to China.

And really going industrial all the way, including in defence, like militarily, the way in which the military industrial complex of of Korea has advanced and is capable of now developing and industrializing its ⁓ over-wing platforms, in part fixed wing platforms, ⁓ intelligent systems. I think it's it's things as c mean meaning C for ISR stuff. It's it's it's quite impressive, like the way in which Samsung Yande and these conglomerates have been able to straddle

the military and civilian and go into digital hardcore. I mean, from car to to to smartphones. Like the alliance between Samsung as a phone and Android is is a very good example of ⁓ Korean knowing where to stop in their and where and how to ally themselves in the pursuit of that. The position of ⁓ Hyundai in high bandwidth memory, which is really critical in the ⁓

Compute in the chip ⁓ packaging of ⁓ AI compute is a good example of that. It's ⁓ it's a small economy that has ⁓ managed to overcome its limitations with a lot of great choices that point to education, to research, to science, to industrial policy, to innovation policy. I mean that that's why I I flag this one as a great example. and that too under the threat of North Korea. Like it's and the the strategic play is not an easy but it's also the

that they have used its potential ⁓ quite beautifully ⁓ there are there are others that ⁓ also speak to more the size obviously a country like brazil because of the size of its population ⁓ remains a very interesting market which is not sufficiently integrated

There there are several others, but to give you two two examples.

Dr Miah Hammond-Errey (16:43)
Excellent. ⁓ let's go to a regular segment. It's called the contest spec contest spectrum. what's a cooperation, competition, or conflict do you see coming in the next 12 months?

Nicolas Miailhe (16:54)
we're seeing US and China, President Qi to be more precise than President Trump seeking a way ⁓ to cooperate or to open a pathway to cooperate on the the systemic risks of ⁓ the race to AGI and how it could wreak havoc in our financial institutions that they that they run and benefit from. I think that that's that's an area of ⁓ of cooperation. conversely, between these two, we have seen the

in a way growing tension and the signalling in Beijing of the the the Taiwan question by President Xi to President Trump in terms of how important it is. And when you look at the various hot spots, strategic or militarily hot spots from from Ukraine to the Strait of Ormuz, to Gaza,

We could see an increasing of the tension in the street of Taiwan depending on how things develop. And it's not necessarily militarily confrontation, but in the case of Taiwan, before the visit of Trump, like President Xi organized the visit of the leader of the opposition of ⁓ of ⁓ of Taiwan in in China to be able to engage into a dialogue to push ⁓ it's it's the Gomitang, the the old enemy or adversary of the Communist Party.

To push an agenda that would clearly ⁓ make any declaration of independence crazy. so I think that's something we should ⁓ we should absolutely look at.

and conflicts. Well the conflict we all know where it is. The threat of Hormuz how do we get out of that? Does it spread? Does it does it contain? Does it subsume? Does it where does it go? It's very unclear at this stage.

Dr Miah Hammond-Errey (18:25)
What are the biggest AI and safety concerns that you have today?

Nicolas Miailhe (18:30)
Well, the velocity shock, the fact that we are moving at a very fast pace towards superintelligent AI without the institutions to prioritize safety over the development, given the techno-scientific ⁓ stack and cocktail that they're composed of. We don't engineer these superintelligence, we breed them. It's more like farming. And so we're in healthcare, we have clinical trials, we have like we have commonsensical ways in which we deal with proaction and precaution.

Over a series of systemic risks in the presence of an understanding of the systemic risks. In AI, we are hypnotized and we are, I feel, running a bit too fast to the cliff. Like ⁓ that's and I'm preoccupied with the way in which we're dealing with this. That's that's ⁓ again the velocity. If we had another 30 years to this potential AGI, and I don't think it's a fundraising scam, I think there is a fundraising element, but but again, it's too late with trillions of dollars poured.

Whether we like it or not, something big is coming out of this. And we're seeing it. The revenue and and trillions of dollars and and and at least one billion of daily users, It's there and now being used in defense, being being early deployed across our critical value chain. so I really ⁓ I really fear that we're not necessarily ⁓ taking into consideration the existential nature.

of ⁓ what's happening in front of us for many good reasons, ⁓ because we need to look after our communities and ourselves and like all these big questions on us will handle them or not.

Dr Miah Hammond-Errey (19:56)
Okay, so then in five years what's the AI conversation we'll regret not having had today?

Nicolas Miailhe (20:02)
Having had much earlier on the conversation of loss of control, strategic loss of control, over the trajectory of development and deployment of these systems. it's not at all the only conversation we should have, of course. And some say that we're having these conversations too much.

Dr Miah Hammond-Errey (20:20)
I'm gonna go to a segment What are some of the interdependencies and vulnerabilities between AI and safety that you wish were better understood?

Nicolas Miailhe (20:28)
Maybe interpretability. so we're dealing with big black boxes, which are ⁓ impossible to ⁓ to explain, and explain in the sense of how they are coming at the inferences they are coming at. Like ⁓ it's it's uninterpretable because of the size of the stochastic mattresses and the number of ⁓ calculation at the level of the weights and the number of weights. I think it's when you combine the number of weights and the number of tokens that have gone into the

⁓ the the pre-training ⁓ and even the post-training ⁓ experiments like it's impossible to to really explain and so opening the the box ⁓ the black box to seek to render ⁓ these ⁓ neural circuits more interpretable to be able to have more control seems to be something like a no-brainer like absolutely a scientific quest of the century to be able to solve this loss of control that I think it's an obvious

It's a double-edged word. Why? Because ⁓ obviously the the way in which gains in interpretability can translate into adding more capabilities versus adding more safety and security, or control rather, control to be more precise, is ⁓ is relative ⁓ to the way in which these ⁓ systems are deployed into the economy for a given desire of controllability. And so it's a necessity, it's not necessary that gains in interpretability which translate into safety.

They will they could improve marginally performance and actually accelerate further the deployment of certain and the development of certain capabilities without translating into gains in safety. And so that's why like strategic techno-scientific pathways to security and control have to be looked and have to be better understood and analyzed with the right framework, which do not exist today, in terms of what's the ⁓ all things being equal and what's the dynamic result of any gains in that.

Dr Miah Hammond-Errey (22:22)
So Prism Eval, one of your companies, has argued that static benchmarks are really broken for evaluating frontier AI. What does rigorous science of measurement for these systems actually look like then? And then who really has the authority to validate it?

Nicolas Miailhe (22:37)
One, don't know what is the the science of measurement for this system is still an unsolved question scientifically and technically, but scientifically. So it's a it's a question in front of you, we should all be humble. now, what we know is that if ⁓ understanding the behavioral dynamics of these artifacts, ⁓ which are, as I said, ⁓ extremely large and stochastic in nature, ⁓ and therefore ⁓ the reproducibility of a given behavior.

or for better or worse is extremely difficult to quantify ⁓ in a ⁓ statistically significant way. And so the reproducibility is something that is that is complicated. And that's what it's pointing to the irrelevance of static benchmarks. Like if you're sending 50 prones pre-written into ⁓ an artifact to qualify a behavior or the absence thereof, like is this serious? Especially if a lot

Of these ⁓ prompts have already been exposed to the public and therefore have leaked into ⁓ the training data set, then it becomes a joke. And and and and I I call that out because of the in intensive ⁓ competitive dynamics and pressure pushing actors to cut corners, including cut metrology, ⁓ metrological corners. So it has to be dynamic because these systems are stochastic and dynamic.

Of a certain nature to qualify any behavior, the way in which behaviors are sampled, and it it translates into the type, ⁓ the the volume ⁓ and the quality of prompts in a given context. The other words I would like to to to the concept I'd like to leave you with is ⁓ good benchmark, good evaluation has also to account for the precision of the context in which we are trying to qualify a given behavior. That's the case for a model, that's also obviously the case for our harness.

And a fully scaffolded agentic system. And that oftentimes is missing. What is useful in the benchmarking concept is that it's also ⁓ risky, in my opinion, to qualify a behavior of one model left alone. It's more ⁓ qualifying and testing a class of model or ⁓ given classes of models to be able to benchmark one versus the other. It's very important to do for the reason I've said before. ⁓ uninterpretable black box, which are

⁓ stochastic in nature.

Dr Miah Hammond-Errey (25:00)
A report earlier this year on AI safety found that while Frontier Labs published their frameworks, which you alluded to earlier, real-world evidence of their effectiveness is limited. Obviously, your position at AISC and Prism Eval gives you a pretty unique read. What's your take on these frameworks?

Nicolas Miailhe (25:19)
Well, they are useful but insufficient because of the velocity to scale, as I said before. And one thing that we should keep in mind as well is that the most ⁓ aggressive ⁓ company on these frameworks, safety wise, which is entropic, has started to renege on ⁓ their responsible scaling policy, confronted with the ⁓ race ⁓ to ⁓ IPO and and to success.

And and their success into that race vis-a-vis OpenAI, they have decided to shed some of the parameters. first, ⁓ they reviewed the responsible scaling policies and then decided to sign line them a little bit. And though that that should be a matter of precu preoccupation. You can be doing washing over them, ⁓ which is one thing, like to say and flash a very important policy and not clearly adhere to it or try to cut corner. I think it's it's it's better than nothing, but that's not good. And you can also decide decide to abandon.

we are speaking a lot these days about ⁓ model specifications, the way in which those who develop these models seek to

Deliver certain characters to these models in how they behave and specify that in order to give more predictability and control. The way in which these model specifications are accounted for besides the system cards or the model cards, which are there to define the with a hopefully high degree of precision, what went into the training, into the development of these models to provide ⁓ readability into the artifact, ⁓ all of that.

needs to needs to move forward in terms of a a driving ⁓ safety framework so in other words when we started a blash lepar defining these ⁓ frontier voluntary commitments the picture in terms of volume and levels of autonomy and capabilities was much more limited than what it is today things have accelerated tremendously and that would demand an acceleration a deepening of these ⁓ frameworks

And also the ability for third party evaluators and backstop regulators to review them. It's not the case. It's really not the case. And that's why AI Safety Institutes, which was, I think, the great outcome of the Bletchley Park AI Safety Summit, are still they are there, but they are weak in nature, underfunded, undercouped in terms of science, under equipped in terms of ⁓ in terms of not necessarily regulation because you don't want them to necessarily be regulators, but they need to serve as a lighthouse for society to understand.

These cocktails of opportunities and risks embodied through an artifact as a frontier safety framework from companies, but insufficient. So that's ⁓ they are really, they should be the ⁓ still the centerpiece, also to be able to inform society and the other mechanisms of accountability from the regulatory in privacy, like there are privacy regulations in many countries, and NAI is hitting them, ⁓ and also ⁓ judges and courts. Like there are a number of cases.

that are percolating across our our courts in child safety and others. ⁓ and so the way in which science can advance as better understanding the cocktails of risk and opportunities is through these frontier AI safety frameworks is really important. Insufficient, but very important. one image to leave you with is they are, in the cocktail of ⁓ regulations or in cocktail of incentives that that drive the behaviors of those who develop, those who deploy, those who invest.

into AI, you want as smart as possible cocktail of self-soft and indeed hard regulatory mechanisms. These pertain to the self-government. And self-government under heavy competitive pressure, we have seen in other domains, is insufficient. And so they need to be coordinated from self-to-soft, meaning standards and code of conduct. But again, it doesn't suffice because of the web of incentives. They need to percolate into

also some regulatory backstops and and that with the level of enforcement that is commensurate like the asymmetry of power between these these frontier firms and the other actors of the accountability chains including regulators is ⁓ is is is complicated like it's it's too it's too imbalanced yeah

Dr Miah Hammond-Errey (29:27)
Can you give me a sense, a couple perhaps, of where commercial incentives are really misaligned with economic security and social outcomes?

Nicolas Miailhe (29:36)
Well, child safety, like the way in which again replaying the story of social media, putting into the hands of ⁓ populations which are potentially vulnerable because of their ⁓ their age, ⁓ into and giving those tools to them because market share is really important. What basically drives the economics of innovation and what investors are looking for and and what they are looking forward to is growth.

Which translates into number of users. And so the way in which you can and you may want, because of the pursuit of return investment, to build those your way through maximizing usage, will land these artifacts into the the brains and hands of people who are not necessarily ⁓ and for sure are not equipped to handle them. Especially when they are not explainable and not easy to control and secure. And that's what leads to ⁓ cyberbullying in the age of generative AI.

Or suicide aided by a general TV. That that's that's a that's a clear case. I think it's ⁓ it's it's happening ⁓ as ⁓ in a way the poster kid of misalignment of incentives and where cutting corners and i is flying into the face of of citizens. There are many other domains of of misalignment. You can talk about workers as well. Like has technology always been there to

Balance out in a ⁓ in a in a in a nuanced way, ⁓ the function of augmentation ⁓ of workers by machines and the functions of substitution of worker by machines. Yes. What is causing a problem now is the velocity at which it's happening and the way in which labour is replaced by capital and how both are compensated for that. It's not at the same level.

Dr Miah Hammond-Errey (31:20)
a it is happening

so fast that those sort of soft and hard mechanisms just are not able to keep up.

Nicolas Miailhe (31:27)
we're seeing a massive displacement of wealth and power. And it's and that's why this question of sovereignty, like is it can should we talk about nation building these days? For sure we should, absolutely. It's still very valid, probably insufficient. That's why we're talking about ⁓ like something global and something of a more humane existential nature. That's why the Pope is appearing as a recourse for those who want to go go beyond

Dr Miah Hammond-Errey (31:31)
Absolutely.

Nicolas Miailhe (31:53)
go deeper into the existential nature or the civilizational nature. I think in between you have a clearly a civilizational moment. I would say that goes beyond nation building.

Dr Miah Hammond-Errey (32:01)
What should we stay focused on, connected to, or grounded in this year to keep sane amid the chaos?

Nicolas Miailhe (32:07)
ground in the in in data. Look at the data and and the evidence that we have. Let's focus to if we are all responding to fear, let's look at evidence and and focus on generating the evidence that we need to be able to drive. Yeah. So data towards evidence is really important.

Dr Miah Hammond-Errey (32:26)
we've got a segment called Emerging Tech for Emerging Leaders and this velocity of change that you've highlighted a number of times obviously requires adaptation at a really quick rate too. What do you see as the biggest shifts for leaders and leaderships from AI?

Nicolas Miailhe (32:40)
As I said again, the velocity at which some of the most fundamental social contracts are being reworked, whether they like it or not. And so it's really like rediscovering the toolbox.

You need to go fast and be inclusive, like to create a critical mass of democratic power to say, Hey, XYZ, this is no no this is yes yes. The the the the leverage, like and the response you have is no, if you want my data center, the copyright data goes. Right? Yes, it goes. And you know it's moving fast, so either we go to you or we go to your competitor, I don't know, Thailand, Malaysia. Sorry, we have to move fast, you know, it's competition. And so

Panic reaction is a natural reaction, but it's really bad. And that's why preparedness and a critical mass of ⁓ support, cross-party, cross-expertise, which is which is really, really hard and needs to be cultivated. Yeah. And so nurturing national conversation. And I think these two can run in parallel, a very technocratic, ⁓ like conversation in policy making, and an eminently political.

conversation. It's very political and aggregating the two at the required level of speed is extremely difficult.

Dr Miah Hammond-Errey (33:51)
It's amazing that we've actually had to deal with incredible existential, global threats before. And we've created entire multilateral institutions. It's almost like we've forgotten we actually can do this. Even though the world has changed and we're doing it in a different way, ⁓ it's sometimes easy to forget that, you know, globally we have confronted some really difficult things before and succeeded.

Nicolas Miailhe (34:12)
Yes. Absolutely.

Dr Miah Hammond-Errey (34:15)
I'll go to my last couple of segments. One is called disconnect. How do you wind down and unplug?

Nicolas Miailhe (34:20)
Number one, I don't sufficiently. Okay. We are part of the problem. This addiction. 'cause it's really an addiction. It it speaks to we speak about child safety, but we don't necessarily speak about the role of p enough. What the role of parents and their own addiction to digital devices and dealing with that. So number one, not sufficiently. Number two, I tend to have a a kind of a radical approach ⁓ which has a name, kitesurfing. I don't kit surf enough, but trust me, to type a message when you're kitesurfing is impossible.

I'm extremely far away from the championship level where that could be I'm sure it's possible, but but really ⁓ it means that you spend your life on the kind to be able to control the kite. So the way I do it is like through sports, through a yeah, and to putting me into situations where I really have to disconnect. And that speaks to the to the level of addiction, to to to the magnitude of the problem. When you have to resort to these kind of radical things, you know it's bad. You know it's if you're a bit self-aware, you're like, Really? This is this is what I need to do.

Dr Miah Hammond-Errey (35:13)
We did we did some research survey for the Cognitive Readiness and Resilience platform, and we found that 99% of respondents had actual strategies to reduce and remove technology addiction dependence because they that they lost track of time completely, had emotional reactions. the way we are interacting with technology is changing and people

are skeptical for good reason because as you say, they actually want to be able to connect with themselves and that cognitive relationship we have ours with ourselves is really important.

Nicolas Miailhe (35:45)
I can fully agree on the routine by the way. I have also a l a few routines. Like I don't watch ⁓ shows. I there are a few things like I know that I block completely to minimize my since I work a lot on screens. So these mini routines as well and this radical choice of certain things, like I'm not even considering. Have you watched that show? No, I don't. Sorry. At least you have to draw the line at some point. It's probably stupid, but that's the way it is.

Dr Miah Hammond-Errey (36:06)
coming up is Eyes and Ears. What have you been reading, listening to or watching lately that might be of interest to my audience?

Nicolas Miailhe (36:12)
I think that maybe one book, which I think is is useful, it's it's it's it's radical, but I think it speaks to the to the essence of the challenge that we have to deal with, by Nate Soros and Helizer Yuskovski called if anyone builds it, everyone dies meaning if everyone builds a type of super intelligence based on today's cocktail and today's competitive dynamics, the risk of ⁓ extinction of humanity is is near certain. That proposition.

Is is is radical enough to well, it it deserves a read for sure. It deserves a read for sure.

Dr Miah Hammond-Errey (36:45)
you've previously said that AI safety is no longer a technical issue, it's a diplomatic one, and you've kind of just expounded on that in in this interview. Diplomacy moves in years and AI capabilities move in weeks. What other governance instruments can match that pace and what can people do to have that confidence?

Nicolas Miailhe (37:04)
As you've said, Mia, there was a time when diplomacy moved in seconds.

We know and our political leaders have already faced in the past situations whereby they have had to react to situations, be it COVID, be it other viruses, be it ⁓ the dealing with the nuclear power. ⁓ I think this is not unknown. ⁓ and I think that we should not panic and and and go back to the fundamentals and seek to accelerate and disrupt marginally these fundamentals. I think it's ⁓ it's really important to not shed

to to find the right balance between institutional innovation from what we have and reinventing something new. I'll give you an example. Do we need a new institution for a real-time ⁓ cryptographic verification certain ⁓ compute operations at the level of pre-training and inference between US and China and others to ensure that when trust is built over certain things that you do not want to build at a given time because it's too dangerous, this trust can be instigated. Does that warrant a new institution?

I think so. should that institution be marked and emerge of the rule of law into form of a treaty, which is a very old institution? Yes, I think so. That's an example of ⁓ not reinventing the will but innovating at the pace that we need. That's I think that's possible. ⁓ that's not the only thing we need, but that's to give you one example of a strategic requirement, that's totally possible and we need to accelerate towards that. And that's still within the realm of of diplomacy.

Dr Miah Hammond-Errey (38:31)
Nico, thank you so much for joining me today. It's been a real pleasure to chat.

Nicolas Miailhe (38:35)
Thanks a lot. Yeah, and appreciate the invitation.

Thanks for listening to Technology and Security. I've been your host, Dr. Mia Hammonderry. This podcast is brought to you by Strat Futures. If there was a moment you enjoyed today or a question you have about the show, send an email to the address in the show notes. Please rate, review and subscribe to help promote technology and security.