Building Post-Quantum Trust for Interconnected AI | ETHSofia 2026

[music]

for clapping. And it's great to have you all here. Thank you.

Amazing. Thank you, Alex. All right, let's hit on. I'm probably, you know, the worst person in terms of crypto knowledge, but let's see how that works. Um Kedo, your presentation was really insightful and I think like when we like what we try to do here is to paint the future of trust, the systems and like stay technical, of course, but like what would the future look like?

And then, you know, try to back propagate to what should we be doing today in terms of systems, in terms of, you know, projects we we work on. Uh because Chain Plus Plus is a showcase of, you know, how to build quantum secure um blockchains even when they're not designed for that. Uh but in practice, when you federate, especially AI stuff, you know, even without a central operator, you still need something to run. And so, I like it's important to think about trust not only in terms of, you know, cryptographic trust, but also like are we you know, are we doing the right the right things on the chain or off the chain? Um and so, what does trust actually originate from in a system like this?

And hopefully, by the end of this panel, we would have a better understanding of what those systems would look like in the future. Um and so, yeah, it's wonderful to be here. Um you know, yes, trust means security, but also means a little bit more than that. Um what are the right problems? Um without further ado, like let's do a round of introductions.

Uh and also, maybe if you can just add let's add add like a small spice to that. Um you know, in one sentence, what is the biggest misconception about post-quantum trust for you. So, if you can just take this and introduce yourself, gentlemen.

Hello everyone. Thank you for joining this conference here and the forum. My name is Peter Nikolov and I am engineer by stove. I have PhD in quantum computing, but most of my career is related with solving hard quantitative and engineering problems and post quantum cryptography is really one of the hardest problems that we solve nowadays, even though we don't know if this is imaginary problem or it's a real problem, because right now we are solving something for the future for the next 10 20 years and uh um

Mhm.

And yeah.

Great.

Kiter, we heard from you, but yeah, let's do another round for everybody who hasn't been at the previous session as an introduction.

Okay. Uh My understanding is that I don't have a crystal ball, so I don't [clears throat] know what the future will be. The best thing that we can do is to prepare to create systems that are flexible and can be adapted. The biggest misconception that I can see in people's eyes for post quantum security is that when you uh choose the best post quantum secure algorithm and implement it, you're done. Reality is just not like this.

Uh it is a journey and exploit will be found in future. Your system needs to be able to adapt. That's probably the most important thing.

Okay, we'll come back to that in a in a second. And Angel, just as an intro.

Yes, my name is Angel Marchev as you heard. I don't like to be titled too much. So, I know that this is a formal thing but anyways. I'm very glad that I'm once again part of the ETH Sofia conference. It's a great place for exchanging ideas.

Now, regarding the misconceptions, I would say I would put two points here. One would be sort of specific. I think there is a clash between the fact that should things go sour the things are going to to to to change very quickly. But, adoption all sorts of uh countermeasures takes a lot of time. So, this is sort of a contradiction that we have to deal with at the beginning to start with.

And my other point is as we are not probably yet living in a post-quantum world, but we are already living in a AI world. Uh a very large misconception is the misconception of super intelligence which is sort of coming and breaks everything and it's not.

Amazing. Amazing. So, let's look at let's look at the future or at least like try to. We don't have crystal balls, but let's try to be as, you know, concrete in a bunch of forecasts uh so that you know, in 5 years we can be really wrong about stuff. All right.

Uh so, Hieu, maybe let's start with you cuz you already, you know, started tackling that that modularity aspect. So, what would Chain Reaction look like in that post-quantum reality? Uh what types of projects would the AI factories be doing, you know, in an ideal world like, based on what we know today. Uh

[clears throat]

Chain Plus Plus' main goal is to enable organizations to share data securely and in a compliant way. We need to live in a post-quantum world. Uh we we are still not there, but we need to prepare for it because according to the the history, you can't just change uh all the cryptography at once when uh quantum computing came. The goal of uh Chain Plus Plus is not a profit. It is funded by EU and by Bulgarian government, basically.

Mhm.

The whole goal and success is measured by the community it builds around it. We need to uh create a community uh that can share data, knowledge, experience, and everything, and eventually start getting uh quicker to the resolutions and facts that

Okay, so

to be

So, for you, it's a stronger community defined by, you know, successful projects. Is there an element to, like, the community being able to withstand certain types of pressure?

Mm not yet. The wider goal of uh Brain Plus Plus Chain Plus Plus is specifically for data exchange. The wider goal for uh Brain Plus Plus is to help organization to adopt AI, uh to give resources in terms of computing, but not only.

Mhm.

Also uh consultancy

Understood.

and uh community and information exchange and everything like this.

Okay, great. Um and Peter, out of all of us, you do have a PhD in quantum algos and um you know how does how do quantum algos reconcile with the post-quantum cryptography world we're trying to describe here? Is there any thing to reconcile actually or not really?

Um actually quantum computing doesn't have anything in common right now with uh post-quantum cryptography. Um post-quantum cryptography is about using classical computers that exist right now to develop new types of algorithms to en- to encrypt data. Uh whether it's hash-based, lattice-based, or any other type of algorithm, it's all classical. It doesn't relate on any in any way with quantum computing. And uh this is one of the paradoxes, if we may say so, that uh uh right now there are I think four or five algorithms approved by NIST to be post-quantum or quantum resistant, but nobody tested the those.

Uh these algorithms uh are even not These are not possible to be tested with classical computers, and we don't have the hardware of uh quantum computers to try to solve such problems like um minimum vector distance or closest vector to to the uh zero zero point or something like this. That's the the the most common uh problem with lattice-based algorithms. And I think uh things will change a lot in the future because um right now the the biggest problem, I think, for encryption and cryptography is harvest now, decrypt later. Because a lot of organizations are harvesting a lot of data which is encrypted. They are trying to decrypt it with certain algorithms, certain ways.

When the quantum computing when the quantum computer is available for you or me to buy for $10 million or for $10,000, we there will be much broader um audience and much much broader perspective of this this task and uh developing algorithms right now for post-quantum world we may have nothing in common with developing algorithms in 10 years or in 15 years because things will change so dramatically that we we don't know how it will look like.

Could you give an example of what that, you know, change would be? Like what like what can we imagine in 15 years to be different once we have the computers to verify?

like if you go back uh 20 years uh or 30 years, uh the main crypto cryptographic algorithm is based on uh prime number factoring.

Mhm.

And Peter Shor developed an algorithm in 1994 which is all um simultaneous algorithm. It's uh using quantum properties but without having a quantum computer. So, nobody tested this algorithm if it's working on a real quantum computer until maybe 10 years ago when the first quantum computer was available in the lab. So, in 10 years I can imagine that someone will develop a new quantum algorithm which is uh way better and solving way different problem than Shor's algorithm because right now not many people are working in quantum algorithms uh field and uh this field will go much larger and larger and uh we may expect uh unexpectable. Amazing.

Uh and Angel, you know, your you have a a unique background both academia but also practice and a very diverse set of different practices and practical projects. One thing you mentioned was, you know, this concept of adversarial situations uh and I think that that unites both AI and uh our topic today. So, can you tell us a little bit like how do we even prepare for, you know, the you know, the the future generations for for those types of challenges whether they AI or or crypto? And how does that affect research? The research that you guys do and the research that, you know, you would be steering.

Yeah. Yeah, thank you. Uh first of all, I would like to continue Peter's thought a little bit and then I'll answer your question.

Amazing.

Exactly. Something that we have talked talked with him multiple times. Uh you should imagine the future of cryptography post-quantum or whatever you call it uh as an arms race. So, there will be people who would be advancing in one side using the current technology, quantum, AI, whatever it is. And there will be people who will be advancing the other side.

So, the so sort of the hacking side and the and the security side. Uh and this is going to go for forever. It's not going to change because we have new new technology. Another um a sort of uh misconception is that now that the quantum quantum computers finally come and start being used um all of our security is gone forever. No, it's not that like this this there is just new people, new algorithms, new ways of thinking.

Uh by the way, the whole topic today basically shows this. So, this is one point. And on on the other back to your question, actually I have a a crystal ball. We don't have generally, but I have a crystal ball. Working with young people, young scholars, young students, currently I know what's going to happen in the next 20 years.

At least I can read it in their eyes in in in a sense. So, yes, it is from my point of view it is important to sort of look what are the current research and educational trends to see what's going to happen in the in the next 20 30 years. And the trends first of all on education are sort of wiffy, I would say. We don't have that many globally. We don't have that many established educational programs.

Maybe several master's programs that are well established. Several even not that many bachelor's. You can say three places or three centers. USA is more on the application side. Europe is more on the research academic side.

China is always there in Asia. Uh but education is I would say it could be much more advanced by this by this point. And by Why I'm saying this? Because this probably shows that the reaction from the market is a little bit slower. Maybe because the market doesn't look at the post quantum or quantum as serious as probably other current topics.

And regarding research, actually research is flourishing right now and probably this conference proves it. We have so many directions of research and you know, various ways. We just spoke in the backroom with some of the today's lecturers. Uh probably the main direction would be lattice lattice-based cryptography and the whole mathematics there. Uh so you can say that one of the pillars would be computer science mathematics.

Another strangely enough is physics. It's not strange that if physics would use something quantum but it is probably original the way they look at it. Now we are working on not not me but people in physics are working on physical data leakage. So this means that you're following a certain processors changing consumption or temperature to decrypt data. Uh and now we are they are working on various methods to sort of hide this to to to spoof this.

Uh there are multiple uh direction. All of them are very original in alternative mathematics. So lattice cryptography we you know at least currently know that there it is virtually unbreakable but there are other always other possibilities to go to. Uh and all sorts of other I would say optimizations of the current algorithm. So the future at least research wise looks bright.

If research likewise it looks bright right now it means that um application wise it will look much different in 10 15 years and this proves just what I just started with and this is it's an arms arms race. One will figure out one way to break the systems and another person or persons or teams or AI will find other ways to to to solve these problems.

Great. So for you you know proof of concepts they will become you know, would be coming faster and everything is going to be coming

this is this is the AI part. Yeah, thank you. I forgot about this this part of the question. Yes, AI is going to basically shorten the production times and you can say and this is not my words, but I like it the way it sounds, the half-life of of implementation is going to be to get shorter.

Okay, amazing. So, now that we know that the future is interesting, um whether that's a good word or bad word, we don't know yet. What should we do about it now? So, how do we design the right trust mechanisms and economics today so that you know, different actors participate in the right way and whether it's in an adversarial setting or not. In other In other words, like what is what should we work on and what is doomed to fail?

Kirill, you know, given your all of your precious knowledge, could you give an example of something concrete that you're working on that would be super helpful?

Yeah. I believe I can give an advice, not an example.

Amazing.

Do the things that you are doing in the last 30 years or so since the era of the internet basically. Catalog your knowledge, your data, understand where your security stands, decide what needs to be secured. So, this is the first step of the things you have to understand what you have and what's important. Second is make of harvest now and decrypt later era. So, someone is probably harvesting your data if it's important.

Uh and third point, third step of the plan, start implementing hybrid algorithms [snorts] where they exist. For example, VPNs or some kind of crypto cryptographic libraries or whatever you're using and needs to. What is don't fail? Choosing just one algorithm that [snorts] works right now and is post-quantum safe. This won't be your last choice.

Definitely.

Right. So, so rotate rotate algorithm equalities, right? And not

Yes.

Okay.

Just make the policies right and everything else will follow.

Understood. And better you know, you're working on state of the the R&D I right now. So, do you think about this future when you do the work that you do today? Does that even enter into the parameter of you know, what is important for for you guys?

Yeah, absolutely. First, I think um related with quantum computing and with cryptography and AI, first AI will go quantum and then cryptography will go quantum for sure. Um even now there are research papers and published materials how quantum computing is possible to speed up the AI development with times, not just with percentage. Uh LLMs are probabilistic models. Quantum computing is also a probabilistic type of computing.

So, it's natural synergy between uh both. About uh cryptography, maybe the next step after the post-quantum cryptography will be quantum cryptography itself. Like, how to develop a cryptography algorithm that uses quantum computing for encryption encryption. Not using quantum computer for hacking or for decrypting existing classical uh cryptography algorithms, but developing the probabilistic nature of cryptography. And I think both worlds of cryptography and AI will develop really fast in uh when the quantum computers will come in, if they come in, because we're not sure yet.

Uh There are many technologies, many research groups, many uh research organizations, um big manufacturing uh companies, and uh uh and all all all sort of organizations that are trying to build a successful quantum computer, a useful quantum computer, but at least of the knowledge of the public mass, nobody did it yet. So,

[sighs and gasps]

when we have it, I think it will speed up things like uh like the internet speeded the the computing in general.

You know, an information an information sharing. So, if I hear you right, you know, quantum computing uh you know, in a way like uh this you don't see this right? You see this as two two parts. Uh you you see this as two parts that are going to converge. We don't know which one is going to get, you know, first uh it you know, at the inflection point.

Maybe it's AI actually because of all of the capex, the trillions of capex in there. Um interesting.

Interesting.

I I see it more as an advantage than a threat.

Mhm.

A threat is if you want to use it in a specific way, but if you want to develop and to to grow, this is an advantage.

Amazing. Um and one final one final question. LLMs and the way the way they're currently implemented in even in in the big the big models, they're very fragile. So, do you think that there's something that we can learn uh from from the research and the practice of quantum cryptography in order to make them more secure because they're probabilistic by nature?

They're very fragile in a certain way. Uh and if you want to escape from this uh fragility, you would need to make um a compensation for something else. So, this is uh point of

Yeah.

whether the one is better than the other and we don't know yet.

No, we don't uh we sure don't and part of what makes them interesting and useful is exactly that fragility. So, uh amazing. And Angel, um you bridge once again theory and business. Um like uh I I think it's it's important to put, you know, all of this, you know, very vague conversation in a in something a little bit more concrete. Uh economic value.

Like, where do you think is the economic value in developing those projects and those mechanisms right now? Um from a cybernetic perspective maybe, uh from a complex system perspective, which is, you know, how you build your research uh career.

Well, yeah, thank you. Uh this is another interesting uh point. Uh so, first of all, I would like to point to point out that uh we shouldn't approach uh post-quantum as a prediction problem. It is more like of another adaptability problem, which means, and this is something that I started on top of our uh introduction. Um when the things happen, they're going to happen quite quickly.

And migration takes quite a lot of time. So, even NIST currently proposes that uh you should start now moving your uh migrating your systems towards the the the new ways, the new algorithms, the new structures. Uh another thing, if you're speaking about uh from the point of view of um of proce- of sorry, cybernetics in this sense. Uh of course, it's a loop. This is why it it takes uh I can translate everything into loop.

This is the point. Uh but uh there should be a shift currently from trusting objects to trusting processes. Uh and this shift, of course, has been sort of reflected in the previous presentation as well. Uh but this is kind of the main from the business perspective point of view. This is because this is my my my point here, you know, these are the engineers here.

I'm Uh I'm the the guy that speaks. So, from the from business perspective, you start migrating, first of all. Second of all, uh as this is a very com- complex, by definition, a complex system, it means that every movable part sort of reflects motions in the other movable parts. It has to be a systematic process. Definitely.

It's not a one-by-one. It's not a patch-up job. And finally, if we if we are to to look at it like again from sort of macro perspective. Since it is an adaptability agility if you wish problem, the quicker you do it, the better competitive advantage you will get.

Amazing. Well, we heard a lot move towards processes like how do you handle adversary situations. You know, the fact that there might be an interesting convergence in practice. Let's give a hand to our panelists. Thank you so much for for this lovely conversation.

[applause] [music]