Author: a16z

Compiled by: Jia Huan, ChainCatcher

AI systems are disrupting an internet originally designed for human scales because they are making collaboration, transactions, and the generation of voice, video, and text cheaper than ever before, and these generated contents are becoming increasingly indistinguishable from human activities. We are already surrounded by CAPTCHAs; and now, we are beginning to see intelligent agents interacting and transacting like humans.

The problem is not the existence of AI; rather, it is that the internet lacks a native way to distinguish between humans and machines while protecting privacy and usability.

This is precisely where blockchain comes in. The idea of ​​how encryption can help build better AI systems (and vice versa) can be nuanced; therefore, this article summarizes several reasons why AI needs blockchain more than ever before.

Increase the cost of AI impersonation

AI can massively fake voices, faces, writing styles, videos, and complete social personas: an actor can impersonate thousands of accounts, opinions, customers, or voters at increasingly lower costs.

These impersonation tactics are nothing new. Any ambitious scammer has always been able to hire voice actors, spoof phone numbers, or send phishing text messages. What's new is the price: large-scale execution of these attacks is becoming increasingly affordable.

At the same time, most online services assume one account corresponds to one person. When this assumption breaks down, everything downstream collapses. Detection-based methods (such as CAPTCHAs) are bound to fail because AI advances faster than the testing methods designed to capture it.

So where does blockchain come into play? Decentralized "proof-of-human" or "proof-of-personhood" systems make it easy for one person to participate, but impersonating many people remains consistently difficult. While scanning an iris and obtaining a World ID may be relatively easy and affordable, obtaining a second one is virtually impossible.

This makes it more difficult for AI to impersonate on a large scale by limiting the supply of IDs and increasing the marginal cost to attackers.

AI can forge content, but encryption technology makes it extremely difficult to forge human uniqueness at low cost. By restoring scarcity at the identity level, blockchain increases the marginal cost of impersonation without increasing the resistance to normal human behavior.

Create a decentralized personality verification system

One way to prove you are human is through a digital ID, which contains everything needed to verify your identity—username, PIN, password, and third-party verification (such as citizenship or creditworthiness) as well as other credentials.

What does encryption add? Decentralization. Any identity system located at the center of the internet becomes a single point of failure. When agents act on behalf of humans—transacting, communicating, and coordinating—whoever controls the identity effectively controls the right to participate. Issuers can revoke access, charge fees, or assist in surveillance.

Decentralization reverses this dynamic: users, rather than platform gatekeepers, control their own identities, making them safer and more censorship-resistant.

Unlike traditional identity systems, decentralized real-person verification mechanisms allow users to control and safeguard their own identities and verify their human identities in a privacy-preserving, trustworthy, and neutral manner.

Create a portable, universal "passport" for intelligent agents.

AI agents do not reside in one place. A single agent may appear in chat applications, email exchanges, phone calls, browser sessions, and APIs. However, there is currently no reliable way to know that these interactions in different contexts refer to the same agent and possess the same state, capabilities, and authorization provided by its "owner."

Furthermore, binding an agent's identity to only one platform or marketplace renders it unusable in other products and all other important places.

A blockchain-based identity layer allows agents to possess portable, universal "passports." These identities can carry references to capabilities, permissions, and payment endpoints and can be resolved anywhere, making agents harder to forge. This will also allow builders to create more useful agents and better user experiences: agents can exist across multiple ecosystems without worrying about being locked into any particular platform.

Achieving machine-scale payments

As AI agents increasingly represent humans in transactions, existing payment systems are becoming a bottleneck. Large-scale agent payments will require new infrastructure, such as micropayment systems capable of handling small transactions across multiple sources.

Many existing blockchain tools—Rollups and L2, AI-native financial institutions, and financial infrastructure protocols—show the potential to solve this problem, enabling near-zero-cost transactions and more granular payment splitting.

Crucially, these tracks support machine-scale transactions that traditional financial systems cannot handle—micropayments, high-frequency interactions, and agent-to-agent business activities.

Nanopayments can be split among multiple data providers, allowing a single user interaction to trigger tiny payments to all contributing sources via automated smart contracts.

Smart contracts allow executable retroactive payments to be triggered by completed transactions, compensating the sources of information that facilitated the purchase decision in a completely transparent and traceable manner after the transaction has occurred.

Blockchain can distribute complex and programmable payment splits, ensuring that income is fairly distributed through rules executed by code rather than centralized decisions, thereby establishing trustless financial relationships among autonomous agents.

Enforcing privacy in AI systems

At the heart of many security systems lies a paradox: the more data they collect to protect users (such as social graphs and biometrics), the easier it becomes for AI to impersonate users.

This is where privacy and security become the same issue. The challenge lies in making the "identity verification" system private by default and obscuring information at every stage to ensure that only humans can generate the information needed to prove that they are human.

Blockchain-based systems, combined with zero-knowledge proof technology, allow users to prove specific facts—PIN codes, ID numbers, eligibility criteria (such as drinking age at a bar)—without revealing underlying data (such as the address on a driver's license).

Applications have received the guarantees they need, while AI systems have been deprived of the raw materials they need to mimic. Privacy is no longer a feature layered on top; it is a core defense.

AI brings cheap economies of scale, but it also makes trust precarious. Blockchain has successfully rebuilt trust by raising the cost of impersonation, protecting human-scale interactions, decentralizing identities, enforcing default privacy, and imposing native economic constraints on intelligent agents.

If we desire an internet where AI agents can operate efficiently without eroding trust, blockchain is by no means a dispensable facility: it is a key piece of the puzzle that enables the healthy functioning of an AI-native internet.