Vitalik's speech in Singapore: AI is simultaneously becoming the biggest opportunity and risk

Author: Vitalik

Compiled by: Wu Says Blockchain

This issue's content comes from a public speech by Ethereum co-founder Vitalik Buterin. The speech took place on October 6, 2026, at the OKX NOW Global Product and Ecosystem Conference in Singapore. Vitalik believes that blockchain technology is far from mature, and AI is simultaneously becoming the greatest opportunity and risk. As modern cryptography, such as zero-knowledge proofs, matures, blockchain can not only achieve "programmable assets" but also "programmable data," allowing users to prove certain facts without disclosing complete information. In an era where AI is massively acquiring and processing personal data, this capability will become increasingly important.

Vitalik believes that AI is likely to become the new user interface. In the future, users may not need to operate blockchain through traditional apps or web pages but can directly let AI agents write scripts, execute on-chain operations, and even understand trading opportunities. This could eliminate some security risks associated with traditional front ends but will also create new issues, such as prompt injection, AI misinterpretation of trading intentions, and whether AI itself is trustworthy. He emphasizes that as AI becomes more adept at discovering software vulnerabilities, higher security standards will not only become possible but necessary.

Looking ahead to the next two years, Vitalik expects Ethereum transactions to be cheaper and more private, and AI will further integrate with on-chain applications. Counterparties may be bots, applications may be written by bots, and even the traditional concept of "applications" may disappear, transforming into different agents interacting directly on-chain. He further suggests that security issues will ultimately extend to the hardware level: "Not your keys, not your coins" may evolve into "Not your silicon, not your keys," and the blockchain industry will continue to be at the forefront of exploring security, privacy, and trusted computing in the entire digital world.

This issue's content comes from a public speech and does not represent Wu Says' views, nor does it constitute any investment advice. Please strictly adhere to local laws and regulations.

Below is the full transcript of the speech:

AI is simultaneously becoming the greatest opportunity and risk

Today, I want to remind everyone of one thing: technology has not stopped evolving, and it is far from mature.

Especially in the past five years, we have seen some very important trends, many of which are being accelerated by AI. AI is simultaneously becoming the greatest opportunity and the greatest risk. I believe we need to understand this, and we also need to understand what new possibilities will arise when we no longer only discuss blockchain, when we no longer only discuss how to put programmable value on the blockchain, but start considering technologies like zero-knowledge proofs.

For example, what would happen if we not only had "programmable assets" but could also have "programmable data"?

From "programmable assets" to "programmable data"

This began roughly ten years ago with the development of zero-knowledge proofs. How many people in the audience have used Zcash? How many have used privacy protocols on Ethereum? And how many have used ZK Rollup? It may still be just a few.

But I expect these technologies to mature rapidly. It is very likely that within two years, most people will have used one of these technologies, even without realizing it.

These are very modern cryptographic technologies. They allow you to prove certain information you possess without revealing complete information. For example, you can prove that certain facts about yourself, an asset, or an object are credible, proving that it meets some condition you care about without having to disclose all information.

Information disclosure should not be an automatic occurrence. You should not have to disclose all information to everyone just to participate in a system. Every piece of information you disclose should be a conscious choice.

I believe that in the AI era, this issue will become increasingly important.

Why privacy becomes more important in the AI era

Rapid technological advancements have brought us many things, but one thing we often lose is privacy.

Now we have very powerful AI agents that can help us think. But if you use Claude to help you think, then Claude can also access your thoughts. For certain information, this is something you do not want to happen; for other information, it may even be legally prohibited for you to disclose.

For example, information involving attorney-client confidentiality or information covered by non-disclosure agreements. In such cases, whether using many mainstream AI products is legal is not entirely clear.

And this is just one example. Similar technologies are entering every corner of our lives at a very fast pace.

You can look around this room and see how many cameras there are. Ten years ago, the number might have been much smaller. Now every person's phone has gone from one camera to three, plus other computers and devices in the room, which may mean there are thousands of cameras here. There are also voice recordings and the collection of various information from the physical environment.

This is just one side of data collection.

AI is enhancing the capability of cyber attacks and vulnerability discovery

From a cybersecurity perspective, we also need to consider how to protect our information and assets.

This year, we have already seen AI truly capable of attacking various systems. We have seen AI escape sandboxes, attack websites, and discover vulnerabilities and exploitable flaws in various software.

At the same time, the Ethereum blockchain has not gone down, nor has Bitcoin.

In the future, higher levels of security will become possible, and a significant part of the reason for this is AI. AI is already helping the Ethereum protocol discover vulnerabilities, assisting us in achieving very complex cryptography, and mathematically proving that these systems indeed have the expected properties.

AI is also beginning to help us with formal verification. Not just at the protocol level, but also at the application level, such as checking whether your assets are secure, whether your data has privacy protection, and various other security issues.

Thus, higher security standards that were previously unattainable will gradually become possible. But at the same time, these standards will also become necessary because as long as there are vulnerabilities in the system, AI will find them. GPT-7 will find them, Claude 7 will find them, DeepSeek 7 will find them. They will all find them.

This means we are experiencing a very interesting and also very challenging transitional period. The way we interact with blockchain, internet services, and AI is already beginning to change rapidly.

AI will become the new user interface

About a month ago, I updated my ENS for the first time without using any traditional interface.

I simply let a locally running AI agent write a script for me to update my ENS hash. The whole process took the AI about five minutes.

I believe this will become the norm. In the future, people will be able to complete complex blockchain operations without using traditional UIs at all.

More and more often, AI will become the new UI.

When AI becomes the new UI, it will bring both benefits and risks.

For example, do you remember the attack on Safe, one of the largest crypto wallets, which ultimately lost about $1.4 billion? That attack was not against the smart contract, nor was it an attack on the code on the chain, but rather an attack on the layer between the blockchain and the users.

If more and more transactions are completed directly by AI in the future, then this layer may disappear. Many of the problems brought by traditional user interfaces may also diminish.

For instance, is the user interface itself secure enough? Is the front end you download every time you open a webpage secure? If North Korean hackers happen to attack the website while you are visiting it, can they take your money as a result? Such risks will decrease.

What new risks will AI operations on the blockchain bring

But at the same time, AI will bring its own problems.

We must confirm whether AI itself is secure. Can AI be subject to prompt injection? Does it truly understand what operations it is performing on the blockchain?

Thus, some problems will be solved, but at the same time, a new set of challenges will arise.

For those developing these technologies, building applications, and everyone participating in this ecosystem, this means both a wealth of opportunities and that we must rethink every layer of interaction with these systems.

The issues we see now and in the future in the blockchain world are not unique to blockchain. The entire internet will ultimately face these issues; it’s just that blockchain will encounter them first.

Why does blockchain encounter them first? The reason is simple: there is a lot of money in the blockchain world.

Attacks that are difficult today may become very easy tomorrow. So the entire internet will eventually have to go through this process, and everything happening in the blockchain world right now is at the forefront.

We are undergoing a very rapid and also very challenging transformation. The capabilities of digital technology may be advancing at ten times the speed of most historical periods.

What will Ethereum look like in two years

Currently, these changes are mainly happening in the digital world, while changes in the physical world are relatively slower. But the physical world itself is also becoming digitized.

Twenty years ago, there might have been only ten cameras in this room; today there may be thousands. In another ten years, there may be five to ten drones flying around here.

So the physical world is also becoming digitized. In the future, other parts of the world will also begin to change.

In this process, we need to ensure the security of protocols and the stability of infrastructure, while not remaining stagnant. We must continuously stay vigilant and consider all the new changes that are emerging: how they affect cybersecurity, how they affect the security of each user account, and how they impact other aspects.

The problems we face are becoming increasingly significant, but the tools to solve these problems are also rapidly strengthening, potentially changing every three months, sometimes even faster.

So, what might Ethereum look like in two years?

You will be able to conduct a large number of transactions, and the transaction costs will be much lower than today, with privacy also being better than today. Applications will no longer just exist for the sake of payments; there will be more and more applications interacting with AI.

In fact, there are already some applications that can help users pay for AI reasoning and can improve privacy when paying for AI reasoning costs. In the future, there will also be applications in which AI itself participates directly.

When transactions, applications, and even counterparties become bots

When you conduct transactions today, your counterparty is likely already a bot, and this situation will only become more common.

Today, your trading counterpart may be a bot, but the application is still written by humans. Two years from now, even the application itself may be written by bots.

In fact, two years from now, there may not even be "applications" as we understand them today. It may just be different bots continuously posting various quotes on-chain, and then your bot will understand all these quotes, tell you what they mean, and judge which quotes are worth accepting.

So many changes will happen here, and these changes will extend far beyond the financial level.

For example, what ensures the security of your personal data? This is an issue for the operating system, an issue for the AI processing this data, and also a hardware issue.

As the physical world becomes increasingly digitized, we must also ensure that the physical world itself is secure.

"Not your silicon, not your keys"

Some people say SI stands for Superintelligence, which means superintelligence. But I believe there is another equally important SI—element number 14 in the periodic table: Silicon.

That is the silicon that makes up all our hardware.

In the past, we would say:

"Not your keys, not your coins."

In the future, we might say:

"Not your silicon, not your keys."

If a system wants to be truly secure, then the hardware that supports this system, as well as the hardware that connects the results of this system to the real world, must also be secure.

Why Blockchain is at the Forefront of Security and Privacy Technology

We must view everything we build as a complete technology stack, considering each layer: all the way up to the user, to the left to the blockchain, to the right to the internet, and all the way down to the hardware.

This is also the area where the blockchain world has always been at the forefront. Today's hardware wallets are likely to become the universal security devices of the future.

So, for all of us, this will still be a very challenging journey, but also very meaningful and interesting.

We need to remember that the blockchain industry will continue to be at the forefront of these issues. These are big problems that the entire world must ultimately face, and many of these problems may have to be solved first by the blockchain world.

I hope everyone continues to think about the challenges brought by these new technologies, as well as the opportunities they present: how we can use them to fend off new threats, and how to make the process of using blockchain more efficient for everyone.

I hope we can not only continue to improve the blockchain world itself but also show the world what security, privacy, efficiency, and fairness can look like, not just in financial systems, but in the entire computing ecosystem.