AI needs Crypto Assets, not TradFi.

Written by: Liu Honglin

In the past few years, AI technology has made rapid advances. Large models, intelligent agents, and automated systems have emerged one after another, transitioning from content generation to code writing, from intelligent customer service to algorithmic trading, AI is gradually moving from a "tool" to an "actor". At the same time, the Web3 field has also begun to enthusiastically discuss the possibilities of "AI + blockchain": using AI to optimize smart contracts, enhance risk control accuracy, assist with on-chain analysis, and so on.

But very few people think in reverse: does AI itself need blockchain?

If we view AI as a participant that gradually detaches from human control and possesses autonomous capabilities, it finds it nearly impossible to operate within the current financial system. This is not an issue of efficiency, but rather a structural problem. The traditional financial system was not designed with machines in mind from the very beginning.

The financial system is designed for "people", while AI is not "people".

The account system is the foundation of the modern financial system. Whether you want to open a bank card, buy a fund, or use payment services, there is one prerequisite you cannot avoid: identity verification. You need to submit your ID card, proof of address, phone number, and may even need to undergo face-to-face video recording to complete the KYC review. The core purpose of these processes is to make the system believe that you are a specific, identifiable, and legally responsible "natural person" or "legal person."

But AI does not belong to either of these categories. It has no nationality, no ID card, no tax number, and does not possess "signature capability" or "legal capacity." AI cannot open bank accounts, cannot register companies, and cannot independently become a counterparty or transaction object. This means it cannot receive money, cannot make payments, and cannot hold assets. In summary: AI is a "non-human ghost" in the current financial system, lacking financial personality.

This is not a philosophical question, but rather the boundaries of a real system.

You let an AI agent buy the usage rights of a server, call an API, or even participate in trading on the secondary market, it must first have a payment method. And any compliant payment method is tied to a "person" or "enterprise" behind it. As long as the AI is not "the subsidiary tool of anyone," but a relatively independent entity, it is destined to be "kept out" of this structure.

Blockchain provides machine-accessible financial protocols.

The biggest difference between blockchain systems and traditional financial systems is that it does not care who you are. You can be a person, a script, a program, or even a "perpetually online" automated agent. As long as you can generate a pair of private keys and an address, you can receive payments, make payments, sign smart contracts, and participate in consensus mechanisms on the chain.

In other words, blockchain is naturally suitable for "non-human users" to participate in economic activities.

For example: An AI model deployed on the blockchain, assuming it obtains data through decentralized storage (such as Arweave) and acquires computing resources from a decentralized computing market (such as Akash), receives payment through a smart contract (settled in stablecoins) after completing the task. This entire process does not require a centralized platform to facilitate, does not need bank card verification, and does not require any human intervention.

This sounds like a futuristic science fiction novel, but in fact, it has already taken shape in some projects. Projects like Fetch.AI, Autonolas, and SingularityNET are exploring how AI Agents can have "economic identities" on the chain, how to provide services for other Agents, and how to autonomously complete transactions and coordination. This form of "machine-to-machine (M2M)" economy has moved from concept to practical testing stage.

AI is no longer a model that relies on humans for input, but rather a self-sustaining entity that can acquire resources, provide services, generate income, and reinvest in itself. It does not require humans to issue paychecks, as it has its own sources of income on the chain.

Why can't the traditional financial system adapt to this scenario?

Because its entire infrastructure is designed around the assumption of 'human behavior'.

The transaction process in traditional payment systems involves someone initiating, someone approving, and someone supervising. The clearing process relies on trust and regulatory coordination between banks. Risk control logic focuses on "who" is doing what, rather than "whether the program is stable." It's hard to imagine an AI wallet opening a bank account through facial recognition, nor can we expect an AI model to complete tax declarations to regulatory authorities.

This leads to all transactions related to "non-human users" in traditional financial systems requiring a "proxy" person or company to operate. This is not only inefficient but also poses a huge liability risk: when AI causes losses, who is liable? When it makes profits, how are taxes collected? These questions have no answers today, but on-chain, at least we have the technical possibility.

Stablecoins: The "hard currency" of the AI world

Many people think that what AI needs is "payment capability," but in fact, what AI needs more is a stable settlement currency. Imagine when an AI agent calls another model or purchases a data API service, it prefers to exchange in "stable value units" rather than highly volatile crypto assets.

This is exactly the significance of stablecoins. USDT, USDC, or future compliant RMB stablecoins provide a financial tool that can circulate freely on-chain while maintaining value stability, serving as the "hard currency" in the world of AI.

Some projects are currently attempting to enable real-time settlement of service calls between AIs using stablecoins, thereby creating a low-friction economic system that does not require "human approval." As the liquidity of on-chain stablecoins increases, AIs can earn revenue directly from tasks and then use this revenue to purchase new service modules or operational resources, forming a truly autonomous machine economy.

Further: The "on-chain legal entity" form of AI?

We can even foresee that in the future, certain AI systems will no longer be dependent on a specific company or research institution, but will exist in the form of a DAO (Decentralized Autonomous Organization) or on-chain protocol.

These AI Agents will have their own funding pools, community governance mechanisms, and on-chain identity systems. They do not require legal registration, nor are they filed in any country, yet they can serve users, receive payments, initiate lawsuits, and publish protocol updates, forming a truly meaningful "digital legal entity" or "AI legal entity."

Their cooperation and competition will be based on smart contracts, mediated by cryptocurrencies, and ordered by on-chain rules. There may be no emotions between them, but there are incentives; no rights and obligations, but there is code execution.

In this process, cryptocurrency is not a speculative asset but rather the underlying protocol of trust between AIs.

Risks and Challenges: We are still far from being ready.

Of course, none of this comes without challenges.

The key custody issues of AI wallets, economic losses caused by model abuse, the verifiability of on-chain identities, the legal eligibility of cross-border AI entities, and the ethical boundaries of algorithmic behavior are all new challenges that must be faced.

More realistically, our existing legal system and regulatory framework provide almost no pathway for "non-human actors." AI cannot sue others, nor can it be sued; it cannot pay taxes, nor can it enjoy property rights; once it is out of control or attacked, who is responsible, and who can be held accountable? All of this requires new legal frameworks, social consensus, and technical governance measures to address.

But at least we have seen a path in some pilot projects - it does not rely on patching up old systems to accommodate AI, but rather on building a more compatible "machine financial infrastructure" to support AI's behaviors.

This infrastructure requires on-chain identity, encrypted accounts, stablecoin payments, smart contract collaboration, and decentralized credit mechanisms. In other words, what it needs is not our traditional "financial system", but Web3.

Written at the end

The development of cryptocurrency initially served those "without accounts," such as individuals, nations, and marginalized industries excluded from the financial system. Now, it may become the only option for "identity-less machines" to participate in economic activities.

If traditional finance is a pyramid built for human society, then blockchain and cryptocurrency might be constructing a "financial foundation prepared for machines."

AI does not necessarily have to possess rights, but it must have operable economic interfaces. And this is precisely the problem that blockchain excels at solving.