Most conversations in crypto focus on price, performance, and security. The common belief is that once the technology becomes secure enough, everything else will fall into place.
But recently, Ethereum co-founder Vitalik Buterin has been bringing the discussion back to a more basic question.
Are these systems actually working in the user’s best interest?
Security is about intent
In his latest X post, Buterin noted that security is about reducing the gap between what users want and what the system actually does.
As he puts it,
“The goal is to minimize the divergence between the user’s intent, and the actual behavior of the system.”
Source: X
That sounds obvious, but it also moves the conversation in an important way. Security today is about understanding that human intent is often impossible to fully encode.
Even a simple action like sending an Ether [ETH] involves hidden assumptions. Is the address correct? Is it really controlled by the right person? If there’s a fork, which version of ETH is the “real” one?
Much of what we rely on falls under common sense. That’s not so easily done by machines, is it? However, there are limits too.
Buterin made sure to note that,
“‘Perfect security’ is impossible… because ‘the user’s intent’ is fundamentally an extremely complex object.”
The internet has the same problem
Buterin sees a similar issue in today’s creator economy. AI can generate unlimited content. Publishing is easy. Execution is easy. The bottleneck is judgment.
“The problem is quality. And so your goal is not incentivizing content, it’s surfacing good content.”
This is the same problem he identified in crypto security. The internet publishes everything; crypto executes everything. Neither inherently understands context.
This is why platforms like Substack stand out to him. Their value comes from elevating the right content through curation, judgment, and alignment.
Increasingly, the challenge is building systems that understand what is worth working on.
This question of intent…
…becomes far more serious in the age of AI. It acts, decides, and interacts on its own.
Buterin recently mentioned that the real issue is not how powerful AI becomes, but who controls it and how its behavior is verified. The goal is to shape its direction.
Warning against systems that concentrate power or operate beyond human oversight, he wrote,
“I want an AI future where we foster human freedom and empowerment…”
Source: X
This is where Ethereum fits in, as infrastructure around it. A neutral layer where AI systems can make payments, prove actions, and build reputation transparently.
Think of this. AI agents could pay for services, interact with other agents, or post security deposits on-chain. These actions create accountability, the behavior of which, becomes traceable. Trust, in turn, becomes verifiable.
As Buterin explained, the goal is to make “trustless and/or private interaction with AIs possible.”
AMBCrypto previously reported that Buterin has proposed personal AI agents to address DAO voting fatigue.
Instead of users manually voting on every proposal, private AI models trained on their preferences could vote on their behalf.
He noted that “the usual solution, delegation, is disempowering,” especially when participation drops and power concentrates with large holders.
As AI expands what machines can do, who ensures they act the way users actually would?
When systems optimize the wrong things
Here’s a different example of the same vein.
- Share
- Tweet
-
