a16z: AI now especially needs encryption technology.

Source: a16z crypto; Compiled by: Jinse Finance

AI systems are disrupting an internet designed for human scale—making coordination, transactions, and the generation of voice, video, and text that are increasingly indistinguishable from human activity cheaper than ever before. We've already suffered from CAPTCHA; now we're starting to see agents interacting and transacting like humans.

The problem isn't the existence of AI, but that the internet lacks a native way to distinguish between humans and machines while preserving privacy and usability.

This is where blockchain comes in. The idea that cryptography can help build better AI systems, and vice versa, can be slightly nuanced; therefore, here we summarize several reasons why AI needs blockchain now more than ever before.

1. Increasing the Cost of AI Impersonation AI can massively forge voices, faces, writing styles, videos, and even entire social personas: one actor can impersonate thousands of accounts, opinions, customers, or voters—and at an increasingly lower cost. These impersonation techniques are not new: any ambitious fraudster has always been able to hire voice actors, forge phone numbers, or send phishing text messages. What's new is the price: the cost of carrying out these attacks on a large scale has become increasingly lower. Meanwhile, most online services assume one account corresponds to one person. When this assumption breaks down, everything downstream collapses. Detection-based methods (such as CAPTCHA) will eventually fail because AI advances faster than the tests designed to catch it. So how can blockchain intervene? Decentralized human proof or identity verification systems make it easy to become "one" participant, but extremely difficult to continuously become "many." For example, obtaining an identity using World ID by scanning an iris may be relatively easy and inexpensive, but obtaining a second one is virtually impossible. This makes it difficult for AI to achieve large-scale impersonation by restricting the supply of identities and increasing the marginal cost for attackers. AI can forge content, but encryption makes it much more difficult to cheaply forge human uniqueness. By restoring scarcity at the identity layer, blockchain increases the marginal cost of impersonation without increasing friction in normal human behavior. 2. Creating a Decentralized System of Personality Verification One way to prove you are human is through a digital identity, which includes everything a person can use to verify their identity—username, PIN, password, third-party authentication (such as citizenship or creditworthiness), and other credentials. What does encryption add? Decentralization. Any identity system at the center of the internet becomes a single point of failure. As agents transact, communicate, and coordinate on behalf of humans, 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 more secure and censorship-resistant. Unlike traditional identity systems, decentralized human verification mechanisms allow users to control and host their own identities, verifying their human identity in a privacy-preserving and trustworthy neutral manner. 3. Creating Portable Universal "Passports" for Agents AI agents do not live in a single place. An agent can appear in chat applications, email threads, phone calls, browser sessions, and APIs. However, there is currently no reliable way to know whether interactions across these contexts refer to the same agent, possessing the same state, capabilities, and permissions authorized by its "owner." Furthermore, binding an agent's identity to only one platform or marketplace renders it unusable in other products and critical scenarios. This fragments the agent experience and makes loading contexts extremely cumbersome. 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 from anywhere, making agents more difficult to forge. This will also enable developers to create more useful agents and better user experiences: agents can exist in multiple ecosystems without worrying about being locked into a particular platform. 4. Enabling Large-Scale Machine Payments As AI agents increasingly act as intermediaries in transactions on behalf of humans, existing payment systems are becoming a bottleneck. Agent-scale payments require new infrastructure—micropayment systems capable of handling tiny transactions from numerous sources. Many existing blockchain-based tools—Rollups and L2, AI-native financial institutions, and financial infrastructure protocols—show promise for solving this problem, enabling near-zero-cost transactions and more granular payment splitting. Crucially, these tracks support machine-scale transactions—micropayments, frequent interactions, and agent-to-agent commercial activities—that traditional financial systems cannot handle. Micropayments can be split among multiple data providers, allowing a single user interaction to trigger micropayments to all contributing sources via automated smart contracts. Smart contracts allow for the triggering of executable retroactive payments after a transaction is completed, compensating sources that contributed information to the purchase decision in a transparent and traceable manner. Blockchain supports complex and programmable payment splitting, ensuring that income is fairly distributed through rules enforced by code rather than centralized decisions, creating trustless financial relationships between autonomous agents. 5. Enforcing Privacy in AI Systems Many security systems suffer from a paradox: the more data they collect (from social graphs to biometric data), the easier it is for AI to mimic users. Here, privacy and security become the same issue. The challenge lies in enabling humans to prove the system's default privacy and hiding information at every stage, ensuring that only humans can generate the information needed to prove their humanity. Blockchain systems combined with zero-knowledge proofs allow users to prove specific facts (PIN codes, ID numbers, eligibility criteria, such as bar drinking age) without revealing underlying data (e.g., the address on a driver's license). Applications gain the necessary conviction, while AI systems are deprived of the original material needed to mimic. Privacy is no longer an add-on, but a core defense. Conclusion AI makes scale cheap, but difficult to trust. Blockchain restores trust, increases the cost of impersonation, preserves human-scale interactions, decentralizes identity, enforces privacy by default, and imbues agents with native economic constraints. If we want an internet where AI agents can operate without undermining trust, blockchain is not an optional infrastructure: it is the missing layer that makes an AI-native internet function.