When we talk about the biggest security risks facing cryptocurrencies, we usually think of one threat: quantum computing.

But Vitalik raises a different possibility that deserves our attention:

What if artificial intelligence manages to accelerate the discovery of new mathematics and finds ways to break cryptographic systems that we consider secure today?

The idea may seem complicated.

Imagine a safe:

Suppose we have an extremely complicated safe. For decades, thousands of mathematicians have tried to find a way to open it without the key.

They haven’t managed it.

Now imagine an artificial intelligence that can do in two years the mathematical work that would normally take 50 years. AI could discover a way to open that safe that humans simply never found.

That is essentially the scenario Vitalik is concerned about.

It doesn’t mean AI can already break Bitcoin, Ethereum, or our wallets. It means we can’t assume a mathematical structure will be secure forever just because no one has found a practical way to break it so far.

What about crypto systems based on lattices?

There is currently a major line of research using something called lattices, or mathematical lattices.

These structures are especially important because they’re considered candidates for protecting systems against future quantum computers. The problem is that we’re assuming the mathematics behind these structures will hold up, but there could be a mathematical discovery we don’t know about yet.

Vitalik raises an interesting possibility:

If AI greatly accelerates mathematical discovery over the next few years, we could discover new techniques for attacking problems that we currently consider extremely difficult.

And this doesn’t necessarily mean lattices will disappear. It could mean we need to use much larger parameters to achieve the same level of security.

What about hashes?

Here’s one of the most interesting parts of his argument.

Hashes have an important characteristic: they’re designed not to depend on a mathematical structure that can be easily exploited. That’s why, when it’s possible to build a system using only hashes, Vitalik considers it a more conservative option in the face of this kind of uncertainty.

In fact, this is one of the reasons Ethereum development has increasingly been exploring hash-based solutions for certain parts of its future architecture.

In very simple terms:

If I can build something secure using hashes, why rely on a more complex mathematical structure that could hide a weakness we don’t know about yet?

This goes far beyond Ethereum, and here’s an important point.

This problem doesn’t affect just cryptocurrencies. Public-key cryptography is used practically everywhere on the internet.

We’re talking about many other things in the world, such as websites, secure messaging, VPNs, Tor, financial systems, identity systems, and cryptocurrencies.

That’s why, if AI really does manage to radically accelerate mathematical progress, the consequences could affect a huge part of our digital infrastructure.

Does this mean we should move our funds?

No. And this part is very important.

Vitalik isn’t recommending that everyone rush to move their funds to new wallets. In fact, he warns that a poorly executed migration could be more dangerous than the problem we’re trying to solve.

If someone makes a mistake during a migration, signs a malicious transaction, or loses their keys, the funds can disappear immediately.

That’s why the recommendation is much more measured:

If it’s easy for you to do so, it may be good practice to keep some funds in addresses that have never been used to make a transaction.

The reason is that certain threats to systems like ECDSA may be different once the public key has been exposed.

But again, it doesn’t mean that the addresses currently in use are unsafe. It means reducing exposure when doing so doesn’t introduce additional risk.

AI or quantum computing?

For a long time, the conversation was: We have to prepare for quantum computers.

Now we might have to add another question: Do we also need to prepare for AI that can accelerate mathematical discoveries?

And what’s interesting is that these two threats are different.

A quantum computer uses the laws of quantum physics to solve certain problems much more efficiently. AI, on the other hand, could help us discover new mathematical methods for attacking problems that we currently believe are difficult.

We don’t know whether this will happen, but that uncertainty is precisely what makes the topic interesting.

Vitalik’s conclusion

The idea isn’t to panic; it’s to think about security for the long term. For Vitalik, hash-based constructions can offer a particularly interesting alternative where possible. And for lattice-based systems, he considers it reasonable to be much more conservative with key sizes and security parameters.

Crypto is still building the infrastructure we’ll probably use for decades, and perhaps one of the biggest challenges isn’t just building systems resistant to quantum computers.

Perhaps we’ll also have to build systems capable of surviving something much harder to predict: an AI that discovers mathematics we don’t even know exists yet.

In this space, we’ll simply never be 100% certain about humans or technology.

DYOR!