a16z's latest research report: Why blockchain is a must-have in the AI era?

Author: a16z

Compiled by: Jiahua, ChainCatcher

AI systems are disrupting an internet originally designed for human scale, as they make collaboration, transactions, and the generation of voice, video, and text unprecedentedly cheap, with generated content increasingly difficult to distinguish from human activity. We are surrounded by CAPTCHAs; now, we are beginning to see agents interact and transact like humans (as reported here).

The issue is not the existence of AI; it is the internet's lack of a native way to distinguish between humans and machines while also protecting privacy and usability.

This is where blockchain comes into play. The idea of how cryptography can help build better AI systems (and vice versa) may be subtle; thus, we summarize several reasons why AI needs blockchain more than ever in this article.

Increasing the Cost of AI Impersonation

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

These impersonation strategies are not new. Any enterprising scammer has always been able to hire voice actors, spoof calls, or send phishing texts. What is new is the price: the large-scale implementation of these attacks is becoming increasingly affordable.

Meanwhile, most online services assume that one account corresponds to one person. When this assumption fails, everything downstream collapses. Detection-based methods (like CAPTCHAs) are bound to fail because the pace of AI advancement outstrips the testing methods designed to catch it.

So where does blockchain fit in? Decentralized "proof-of-human" or "proof-of-personhood" systems make it easy for one person to participate, but difficult to impersonate many. While scanning an iris and obtaining a World ID may be relatively easy and affordable, obtaining a second one is nearly impossible.

This makes it harder for AI to achieve large-scale impersonation by limiting the supply of IDs and increasing the marginal cost for attackers.

AI can forge content, but cryptography makes low-cost forgery of human uniqueness increasingly difficult. By restoring scarcity at the identity layer, blockchain raises the marginal cost of impersonation without increasing the friction of normal human behavior.

Creating Decentralized Proof of Personhood Systems

One way to prove you are human is through a digital ID that contains everything needed to verify identity—username, PIN, password, and third-party proofs (like citizenship or creditworthiness) along with other credentials.

What does cryptography add? Decentralization. Any identity system centralized on 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 participation rights. Issuers can revoke access, impose fees, or assist in surveillance.

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

Unlike traditional identity systems, decentralized proof-of-human mechanisms allow users to control and safeguard their identities while verifying their human identity in a privacy-preserving and credibly neutral manner.

Creating Portable Universal "Passports" for Agents

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

Moreover, binding an agent's identity to just one platform or marketplace makes it unusable in other products and all other important places.

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

Enabling Machine-Scale Payments

As AI agents increasingly transact on behalf of humans, existing payment systems become bottlenecks. Large-scale agent payments will require new infrastructure, such as micropayment systems capable of handling tiny transactions across multiple sources.

Many existing blockchain tools—Rollups and L2s, AI-native financial institutions, and financial infrastructure protocols—show potential for solving this issue, enabling near-zero-cost transactions and more granular payment splits.

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

Nano-payments can be split among multiple data providers, allowing individual user interactions to trigger small payments to all contributing sources through automated smart contracts.

Smart contracts allow executable retroactive payments triggered by completed transactions, compensating those information sources that facilitated purchasing decisions in a fully transparent and traceable manner after the transaction occurs.

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

Enforcing Privacy in AI Systems

At the core 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 is to make "proof of personhood" systems default to privacy and obfuscate information at every step, ensuring that only humans can generate the information needed to prove their humanity.

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

Applications get the assurances they need, while AI systems are deprived of the raw materials necessary for imitation. Privacy is no longer an add-on feature; it is a core defense.

AI brings cheap scale effects but puts trust at risk. Blockchain successfully rebuilds trust by raising the cost of impersonation, safeguarding human-scale interactions, decentralizing identity, enforcing default privacy, and providing agents with native economic constraints.

If we desire an internet where AI agents can operate efficiently without eroding trust, blockchain is not a nice-to-have facility: it is a key piece of the puzzle for a well-functioning AI-native internet.