Ethereum co-founder Vitalik Buterin said AI is becoming both the biggest opportunity and the biggest risk for the crypto sector, and argued that blockchains will likely be the first systems forced to deal with the security consequences at scale.
Buterin made the remarks during a public speech delivered on Oct. 6, 2026, at the OKX NOW Global Product and Ecosystem Conference in Singapore. The speech, translated and published by Wu Blockchain, laid out his view that blockchain technology is still far from mature, while AI and modern cryptography are reshaping what onchain systems may become.
In Buterin’s view, the conversation can no longer stop at putting programmable value onchain. With zero-knowledge proofs and related techniques maturing, blockchains can move from “programmable assets” toward “programmable data,” giving users a way to prove facts without revealing all of the underlying information.
From programmable assets to programmable data
Buterin said the shift traces back roughly a decade to the development of zero-knowledge proofs. He pointed to Zcash, privacy protocols on Ethereum, and ZK rollups, saying the number of people who have actively used those tools may still be small today, but he expects that to change quickly.
He said it is very possible that within two years, most people will have used one of those technologies without even realizing it.
He described them as modern cryptographic tools that let a person prove possession of certain information without disclosing the full data set. A user, in that model, can prove something about themselves, an asset, or another object, and show that a condition is satisfied without exposing everything.
For Buterin, disclosure should not happen automatically. People should not have to reveal all information to everyone just to participate in a system. Each disclosure, he said, should be a deliberate choice.
He added that this becomes more important in an AI era.
Why privacy becomes sharper in an AI-heavy world
Buterin said rapid technical progress creates many benefits, but privacy is often one of the things people lose along the way. Strong AI agents can help people think through problems, he said, but if someone uses Claude to help with thinking, then Claude also gains access to those thoughts.
For some categories of information, that is not acceptable. For others, he said, disclosure may not even be legally allowed. He cited attorney-client privilege and information covered by non-disclosure agreements as examples, adding that it is not fully clear whether the use of many mainstream AI products is lawful in every such case.
He framed that as only one example of a much broader trend. Similar technologies, he said, are moving into every corner of daily life at high speed.
Pointing to the room around him, Buterin asked the audience to consider how many cameras were present. Ten years ago, he said, the number would have been much smaller. Now each phone may have three cameras, and when combined with computers and venue equipment, a room like that might contain thousands of cameras, along with voice recording and other forms of physical-environment data collection.
That, he said, is only one side of the data-collection story.
AI is getting better at attacking systems and finding flaws
On cybersecurity, Buterin said the industry also has to think about how to protect information and assets as AI capabilities advance. He said 2026 has already produced signs that AI is beginning to show real ability to attack different kinds of systems. He pointed to cases where AI escaped sandboxes, attacked websites, and found bugs and exploitable weaknesses across software.
At the same time, he noted that Ethereum has not gone down, and Bitcoin has not gone down.
He argued that stronger security standards are becoming achievable, and that AI itself is one of the main reasons why. AI is already helping Ethereum identify protocol vulnerabilities, helping developers implement very complex cryptography, and helping prove mathematically that systems have the properties they are supposed to have.
He also said AI is beginning to support formal verification not only at the protocol layer, but also at the application layer, including checks on whether assets are safe, whether data remains private, and whether other security conditions hold.
That changes the bar. Standards that used to be out of reach may soon become attainable, he said, but they will also become mandatory because if a system has a vulnerability, AI will find it. In his words, GPT-7 will find it, Claude 7 will find it, and DeepSeek 7 will find it.
He described the result as a transition period that is both fascinating and difficult, with the ways people interact with blockchains, internet services, and AI already changing quickly.
AI as the new user interface
Buterin gave a personal example involving ENS. About a month before the talk, he said, he updated his ENS for the first time without using a traditional interface at all.
Instead, he asked a locally running AI agent to write a script that updated his ENS hash. The whole job took the AI about five minutes.
He said he expects that pattern to become normal. In the future, users may be able to complete complex blockchain operations without touching a traditional UI, because AI will increasingly serve as the UI.
That shift, in his telling, cuts both ways. It can remove some existing weaknesses while creating new ones.
Some front-end risks may fall, but new AI risks appear
Buterin referred to the Safe attack, describing Safe as one of the largest crypto wallets and saying the incident ultimately led to roughly $1.4 billion in losses. He said the attack did not target smart contracts or onchain code, but instead the layer between the blockchain and the user.
If more and more transactions are completed directly by AI, he said, that layer may shrink or disappear, and some long-standing interface risks could ease. He listed examples such as whether the UI itself is secure, whether the front end re-downloaded every time a user opens a webpage is secure, and whether attackers, including North Korean hackers, could strike a site at the moment of access and use that route to take funds.
But he was equally clear that AI introduces its own risks. The industry, he said, has to establish whether the AI itself is secure, whether it can be hit by prompt injection, and whether it actually understands the blockchain actions it is carrying out.
So some problems may be solved, but another set of problems arrives in parallel. For developers, app builders, and everyone involved in the ecosystem, that means opportunity, but it also means rethinking every layer of interaction with these systems.
Why blockchains will face these issues first
Buterin said the problems appearing in crypto are not unique to crypto. The wider internet will run into them too. The difference, he said, is that blockchains will get there first.
His reason was simple: there is a lot of money in crypto. Attacks that are hard today may become easy tomorrow. In that sense, he said, what is happening in blockchain systems now is the front line of a transition the internet as a whole will eventually have to go through.
He described digital technology as advancing at what could be ten times the pace seen in most earlier periods of history.
What Ethereum may look like in two years
Buterin said most of the visible change is still happening in the digital world, while the physical world is moving more slowly. Even so, the physical world is being digitized too.
He again used cameras as an example. Twenty years ago, a room like the one he was speaking in might have had 10 cameras. Today it might have thousands. In another 10 years, he said, it might even have five to 10 drones flying around inside.
As that process unfolds, he said, the sector has to keep protocols secure and infrastructure stable without freezing in place. It also has to keep reassessing how every new change affects cybersecurity, user-account safety, and other parts of the stack.
The problems are getting bigger, he said, but the tools for solving them are getting stronger just as fast, with meaningful shifts sometimes arriving every three months, and occasionally even faster.
Looking specifically at Ethereum over the next two years, Buterin said users should be able to carry out many transactions at costs that are much lower than today, with stronger privacy than today as well. Applications will not be limited to payments for their own sake, he said, and more products will emerge around interaction with AI.
He added that some applications already help users pay for AI inference and improve privacy around those payments. In the future, he said, some applications may include AI itself as a direct participant.
Counterparties may be bots, and apps may be written by bots
Buterin said that when people trade today, the counterparty is often already a bot, and that trend is only likely to deepen.
Right now, he said, the counterparty may be a bot while the app is still written by a human. Two years from now, the app itself may be written by a bot. And beyond that, what people now think of as an “app” may no longer exist in the same form.
He described a possible setup in which different bots continuously post quotes onchain, while a user’s own bot interprets those quotes, explains what they mean, and decides which ones are worth accepting.
The implications go beyond finance, he said. The safety of personal data depends not only on the operating system and the AI processing that data, but also on the hardware beneath it. As the physical world becomes more digital, the security of the physical world itself has to be part of the discussion.
“Not your silicon, not your keys”
Near the end of the speech, Buterin turned to the term “SI.” Some people use it to mean superintelligence, he said, but another SI matters just as much: silicon, element number 14 on the periodic table.
From there he proposed a new line: “Not your silicon, not your keys.”
He presented it as an extension of the well-known crypto phrase “Not your keys, not your coins.” If a system is going to be truly secure, he said, then the hardware that runs it, and the hardware that connects its outputs back to the real world, also has to be secure.
Blockchain stays at the edge of security and privacy work
Buterin said builders need to see what they are making as one complete technology stack and think through every layer of it: upward to the user, leftward to the blockchain, rightward to the internet, and downward to the hardware.
That is one reason blockchain remains at the front edge of these questions, he said. Hardware wallets used today may end up becoming more general-purpose security devices in the future.
He closed by saying the blockchain industry will continue to sit at the front line of questions around security, privacy, and trusted computing, because these are issues the whole world will eventually have to face. In many cases, he said, crypto may have to solve them first.
The speech was a public presentation and does not represent the view of Wu Blockchain. It was published with a notice that it does not constitute investment advice.
