We're going to kick things right off by our very first fireside chat of the today. So, without further ado, please welcome Charles from Ledger.
[applause]
Welcome. Good morning, everyone. I think it was perfectly fitting that we bring in the best French company we know in this space at this first event and kick things off. And not only that, a product that all of us have been using for almost a decade now. Um I have a few questions that I want to dig into, but the thing I want to start off with is understanding kind of how your role role has changed.
You were previously the Chief Security Officer at Ledger. You recently became the CPO, and I'm curious to know how your day-to-day has evolved. Are you thinking more about the product? Do you think about things differently? What's been same?
What's been different? Yeah, sure. My my background is really cryptography and and security. Those are topics I'm interested in since I'm 12. When I got a computer, the first thing I did was to try to hack everything.
That was easy at the time. Um and then I worked in the security area for my whole career. And right before Ledger, I was heading a security evaluation lab. And basically, we were evaluating the security of different devices for banking, identity, HSM, smart card technology, this kind of thing. Um funnily enough, like with a banking card or a passport, we are at Ledger we are using the same technology, but the main difference is that in the banking card or in the passport, the secret belong to the bank or to the state.
While with a Ledger, they belong the the secret belong to you. When I joined Ledger in 2017, I created the security department of Ledger that we call today the the dungeon. Um and basically, the idea of this team is to raise the bar for security when it comes to Ledger product, but also in the whole ecosystem. And when you when you are in security, there is something that you know for sure that the best way to make secure product is to try to break them. And this is basically what what they do day-to-day.
They try to break our product. And when they find vulnerabilities, they work with the engineering team in order to improve things. Um and in 2019, so now it's a little bit a long time ago, I moved to moved from C Chief Security Officer to Chief Technology Officer. Um and I agree it's it's a little different because instead of trying to break our products, you're we are trying to make great product and make them secure. But at the end, security is really the core value proposition proposition of what Ledger is doing.
So, this move from Chief Security Officer to Chief Technology Officer was was quite natural in the end. That's awesome. I mean, the world was a lot more different in 2017 and even 2019. I mean, even L2s were not really out there. So, one thing I'm curious about is understanding how over the last 7 to 9 years have things changed on the way you think about custody?
Are are things a lot easier now cuz you've learned so many things and now you can support every blockchain? Or does that mean now things are harder to secure cuz the surface area is much much bigger? Yeah, as you said, the the space has changed significantly in the in the past years. Um in 2017, the ecosystem was mostly about Bitcoin, and that was the beginning of of Ethereum. Ethereum was created a few years before, but the ecosystem was really still very small in 2017.
Um but luckily, what the technology that we are using was was future-proof. Because what we have built is really like technology platform that that put the security front and center, but this is more like a technology platform. We have built an operating system, and this operating system allows anyone to build his own application in order to support his own security problem. So, all the blockchain that popped up in the in the past years were quite easy for us to integrate because we have this operating system, and anyone can build his own app to support his new blockchain. So, this move was really a good one, and it really put it it really put Ledger in a great position in order to support the the new protocol and and blockchain that appeared through the years.
And and this this period was especially exciting because in terms of technology, plenty of progress were made. If you think about the 2017-2018 era, the blockchain was were very slow. Bitcoin didn't change much, but the Ethereum space and the new space that were created have evolved significantly. And now we have a scalable blockchain that enables plenty of new use case, new protocols. Um and and now we are in this point in time where traditional finance think that this technology is really suitable for their needs.
And this is what we are trying to see what we are starting to see. Traditional finance finally understanding that the blockchain technology is far better than anything that they already use because it's faster, it's cheaper, it's natural, and it's you have you also have composability and so on. And we are in this point in time where all the progress that we made in the technology enables new use case, and the biggest one today today is definitely finance. Amazing. I I I think that also sets the stage for like how much in a way perturbations there are going to be in the in the future.
So, the thing I want to turn into, which is kind of one of the the key discussions for this chat, is understanding how the world looks like post kind of quantum cryptography here. So, my kind of first question to you is how real do you think this is and how close are we to what the actual problem is on breaking all of cryptography? Yeah, it's it's a great question. Um So, just to give a little bit of context, um we we are talking about post-quantum cryptography because there is this idea of creating quantum computers. And quantum computers theoretically could break cryptography.
Uh quantum computers are very different computers uh compared to classical ones. They are leveraging quantum phenomenon, and they are using qubits and so on. So, very different. With a quantum computer, you cannot solve general-purpose problem. They won't replace classical computer.
But they are very good at a few things, including physics, uh chemistry, but also breaking cryptography, unfortunately. Um we like humanity discovered the idea of quantum computer in the '80s, '90s. And at the time, we theoretically invented some algorithm that can break cryptography, notably two. One is Shor's algorithm, and the other one is Grover's. As it's not a technical conference, I won't enter into the details.
Uh but we know theoretically that with large enough and reliable enough quantum computer, it's possible to break cryptography with these two algorithm. And I'm in the cryptography world for around 20 years, and I hear from the time It's always been 10 years away, right? That in 5 years we will have a quantum computer capable of breaking cryptography. Um however, so I I I say this to be a little bit picky. Uh however, we are seeing like specific progress recently.
And we are seeing like two two kind of progress. On one side, we are improving the theoretical algorithm. Like Shor and Grover are improved. The recent paper that Google announced, I think that was 2 days ago.
2 days ago. Uh was exactly doing this. They said that it's finally possible to improve Shor's algorithm so that you don't require um a very large number of qubits. So, they they decrease the number of qubits necessary. And on the other side, we create quantum computer with a larger number of qubits.
However, we are still far from bridging this gap where you will have enough quantum qubits where qubits in order to solve cryptographic problems. However, I think the question is not really to know when this quantum computer will happen. This is really a moot point. The real Cryptography is useful in order to create trust into the security of our system. This is really what cryptography is about, creating this trust.
As soon as this this trust is awarded, then we need to find solutions. And the good thing is that we have the tools. This is called post-quantum cryptography. And this this kind of cryptography is immune to quantum computer attacks. So, now it's really a matter of migrating all the system to this post-quantum cryptography.
The good thing is that we have the NIST, which is the National Institute of Standard and Technology. It's it's a US one that standardized plenty of things, including cryptography. And they have created standards for this new cryptography. And they also created a timeline where all the industry need to move towards these new standards. So, they mandate to the move towards post-quantum cryptography by 2030 and 2035, depending on on the kind of application.
Now, it's really a matter of doing this move. However, there are so many systems like think about telecommunication, the internet, defense, finance, everything is using cryptography. So, we have like 9 years, but 9 years is very short to move towards post-quantum cryptography. So, we really need to move fast and and start today. And for blockchain, the issue is mostly to reach the social consensus.
Because when you are a company, when Google said they will move towards post-quantum cryptography by 2029, they just have to decide it and to do it. When it comes to find the social consensus for Bitcoin or Ethereum, it's much more complicated because there is no central entity, there is no CEO of Bitcoin or no Ethereum that will say, "Okay, we move towards post-quantum cryptography by 2030 and we will do use this algorithm, this algorithm, and this one." So, today we we are really in this point in time where we kind of agree that this move is necessary. And now it's really a matter of agreeing on the timeline and on the how. And there are plenty of technicalities on which kind of signature we would like to use and this kind of thing.
And at Ledger, what we are doing is to participate into this discussion, drive a little bit the choice, and also we are starting the implementation of this post-quantum cryptography algorithm within our operating system so that the application will be able to leverage the the secure hardware to to to use post-quantum cryptography and sign transaction into a post-quantum world. There's a lot of stuff in there. So, I have a couple of follow-ups to a few things. So, one in the one of the pieces for everybody in the audience 2 days ago when Google announced that, "Hey, we should move the their own timelines to 2029." That's That's 3 years from now.
Or 2 and 1/2 actually. Um They basically said that instead of needing a million qubits, you can get this done by 41,000 qubits and that's kind of where a lot of the the timeline acceleration comes from. Was this from their side an attempt to say that this is generally for a specific set of curves that this would break or they were kind of commenting on just their own systems. Maybe that's a quick clarification there and then a few follow-ups. Yep.
So, Shor's Shor's algorithm is very generic for a large class of math problem and at the beginning we were talking about RSA, which is the ancestor of elliptic curve cryptography. And we know that with Shor we can break all this class of problems and it was very well studied for RSA. And what Google did was to study Shor implementation specifically for elliptic curve cryptography. And again, it can it can break any kind of curve as long as it use elliptic curve cryptography. Awesome.
Second question. So, if if I understood what your comment was about blockchains here, you're saying that largely there's a lot more drama in the Bitcoin world than there is on Ethereum on this, but largely we are just debating the timelines in which implementations to go for, but there's not unsolved problems on which curve to use cuz this is this exists, there are standards. We just have to get this in the codebase. Yeah, there are standards, but there is no one standard. Which is the best thing about standards from what I hear.
Exactly. So, there are several standards with pros and cons. You have like hash-based post-quantum cryptography, which is we like the community like like
blockchains are set up now. We we just have most blockchains are set up now with electric curves. But but not for for signature. But it's possible to use hashes to create post-quantum signatures. Got it.
And so, this is a class of signature that are post-quantum cryptography and the community likes them because we have it's it's well studied, we have stronger guarantees. However, you have plenty of cons like the signatures are large, it takes more time to compute them. So, you have you have this kind of drawback. And you have another class of post-quantum signature, which is more lattice-based. So, it's another class of math problem and this one is a little less studied than hashes.
Frankly, it's totally secure. The NIST standardized them, so we have good trust that it's secure. And you have plenty of benefits like the the computation is less heavy, the signature are smaller. And for blockchain, I think it's it's also important because
that. Yeah, you don't want to have very very large signature, otherwise the block will be huge, especially for Bitcoin where you might you might remember the block was. So, today this is where we are. The debate is ongoing, but I think we should settle the debate as soon as possible and and start the migration timeline because it will take time. So, one other side effect from what I just understood seems like for both Bitcoin and Ethereum, if you were to implement these post-quantum curves, the addresses will change.
Is that correct or or how do you or are you able to maintain compatibility or does that require a full change of everybody's addresses and and private keys? Yeah, the the transition will be a little bit tricky. There are there are plenty of ways. If if we had like tomorrow a post-quantum computer and we and we have to migrate everything very fast, for Ethereum we have a solution. The idea is to create a smart contract where a spending rule would be will be the verification of a post-quantum signature.
So, that you don't change anything on the chain and you just create smart wallet that are gated by a post-quantum signature. So, it's it's a little bit hacky, but it works. Fortunately, we won't have like quantum computer tomorrow. We have at least 5 years from what I hear. Yeah, so we have we have time.
So, just in case we have this this possibility. Otherwise, migrating everything means creating new addresses, moving your funds from your vulnerable address to these new addresses. That might require also a change in the size of the blocks, especially for for for Bitcoin. Also, there are questions about what do you do for coins that are never migrated? Do you say, "We let them here and we wait for someone to steal them with a quantum computer?"
Which would be a big problem because that that would create a like a large selling pressure on on this coin. Or do you say, "Okay, we will freeze them and then we will pay limit the same amount." It's another possibility, but it would change the the the the supply.
So, there are plenty of that question. There are no like perfect solution and this is really what we need to talk about now, decide all of this question, settle them, and migrate. This is this is this should be the main discussion today. Amazing. Um I do want to switch a little bit on the topics and I want to understand how you as Ledger think about AI agents and payments.
Um There's a I mean, a lot of money is self-custodied in crypto. What does it mean to have a an opinion and play in this in this world? Yeah, sure. Uh there are plenty of things to say about AI. I will start with a security angle.
Um AI is changing everything when it comes to security. Like security is about creating an asymmetry between the attackers and the defense. And the idea is to make sure that finding a vulnerability and exploiting them would be much more expensive than what you would extract from the secret, from whatever you would able to break. And this is really the main paradigm for security. And this is completely changing with AI because the cost of finding vulnerabilities and exploiting them is tending to zero.
So, now we are in this tricky moment where attackers have a big advantage against defense. There is plenty of hope because we we can use also AI, leverage AI to build more secure system using like formal verification, zero-knowledge proof, secure enclave. So, there is some hope, but now the the transition into this more AI-oriented world would would be a little bit tricky. And when it comes to agents, yes, we are talking a lot about the idea that your agent could buy your airplane tickets or or whatever. Um and definitely agents will use crypto.
Crypto is more efficient than anything anything else. And agents don't really care about the UX of Visa or or MasterCard. An agent would just want to be as efficient as possible when it comes to payment. So, if you don't specify anything, your agent will use crypto to to do any kind of payment. So, blockchain will be the rail for agentic payment.
There is no debate on that. However, you don't really want to give your seed and your secret to to your agent. An agent is not really secure by design and and also if you give it your full money, it might do
That's not a it's going to be a not a good story. Yeah, it's a very bad idea. There are already plenty of horror story around this. So, what we are thinking about at Ledger is is more to delegate delegating rights to the agent. And the idea would be to create policies that could be enforced by hardware where you will say that my agent can spend up to $10 a day without my consent.
And above this threshold, then I would be prompted on my device so that I can I can consent for this transaction. And and and for other type of transaction, don't even propose me because for sure I will decline them. So, the idea is really to create fine-grained permission and the governance engine so that we give a safe playground for the agent to use the money and delegate some rights. We already have something similar on the B2B side. We have a ledger enterprise solution that enables to do to implement this governance.
And as of today, you can directly plug your agent to our MCP or API and create the governance rules that will enforce the use case I just mentioned. So, it's already possible on the on the B2B side. And we are thinking about how to enable the same kind of thing on the B2C side. Amazing. Well, my kind of last question to you is there's a lot of opportunity right now in AI, there's a lot of just general security related things that we know we have to do in the next few years.
And you also kind of recently talked about Donjon, your research lab. So, where is most of your time going now and kind of what are important things to you for the next 2 years? I think we covered some of those topics. One of the big point of attention for me is really this shift that is provided by AI when it comes to security. I think like a lot of systems will be will be broken in the coming months and years.
So, collectively we really need to raise the bar for for security. Like use formal verification, use zero-knowledge proof, use secure enclave. Like now, security is not an option anymore. Like for for people who are using like software wallet with a seed in their phone, we all know that's not secure. But now, you will be you will be wrecked.
There is no doubt about that. There is it's it becomes so easy to to break so many different system and operating system, Android, iOS. So, that it it will just happen. And we are already seeing such large-scale attack on the field. So, we collectively need to raise the bar for security.
So, I'm really we are really focusing on on on those topics, enabling agentic payments in a secure manner with a nice governance and policy engine framework and leveraging those new technologies such as like ZK and formal verification. Amazing. Charles, thank you so much for this amazing conversation. Please give him a round of applause. Thank you.
Automatic transcript — names and jargon may be misspelled.