Trust in AI
Over the past decade, AI has been defined as a form of "intelligent productivity"—assisting humans in decision-making, optimizing processes, and reducing costs. Today, however, a more fundamental question is emerging:When AI no longer merely thinks and expresses but can autonomously "spend money," how will the economic order of this world be rewritten?
The advent of AI payments signifies the first time machines participate in value exchange. AI transforms from a cognitive system into an economic entity. It requires identity, accounts, credit, contracts, clearing, and even liability. As we discuss the next steps for AI, the focus shifts away from algorithms, computing power, or the scale of large models to more foundational questions: When AI truly participates in economic activities, how will it "pay"? Who grants it accounts, credit, clearing capabilities, and liability?
The redesign of payment systems will determine the boundaries of the intelligent economy.We have identified three distinct paths emerging around the question of "how AI pays."
One isan institutional path centered on identity and regulation, attempting to enable AI to operate legally within the existing financial system; another isa crypto-native path built on blockchain infrastructure, allowing AI to become autonomous economic nodes; and the third isa vertical integration path centered on corporate ecosystems, facilitating the initial implementation of AI payments through productization and encapsulation.
These three routes resemble the civilizations in The Three-Body Problem, independent yet interacting. They are not mutually exclusive; rather, under different logics of trust, they explore the same question:In a world where machines can also fulfill contracts, how should trust be rebuilt?
Google: Enabling AI to "Spend on Behalf of Humans" Within the System
Google’s approach most closely resembles imperial thinking. Its AP2 (Authorized Payment Protocol) is an "AI Authorized Payment Protocol" designed toembed machine economic behaviorinto the existing financial system. In other words, it does not aim to give AI its own accounts, but rather to allow AI toexecute payments on behalf of humans within authorized limits.
Its core mechanism comprises three components:identity verification, authorization credentials, and payment execution.
Users first define their AI permissions within the system—for example, the AI may pay subscription fees, place advertisements, or settle computing bills on their behalf, but cannot transfer funds or make investments. Whenever the AI initiates a payment, the system generates a short-term cryptographic authorization (similar to a one-time digital signature). After Google’s network verifies the identity, the transaction is executed. Upon completion of the payment, the authorization automatically expires.
Under this mechanism,AI acts merely as an executor, not as the account holder.The funds spent still originate from human wallets, and clearing continues through banks, card organizations, or Google Pay payment channels. AI serves only as an "agent," with its permissions and boundaries jointly defined by users and regulators.
The problem Google seeks to solve is:How can AI safely help humans spend money without exceeding authority or violating laws?It is not concerned with freedom, but rather withcontrollability and compliance. This represents "imperial trust": trust derived from identity and regulation, rather than from algorithms.
Coinbase: Enabling AI to "Spend Its Own Money"
Coinbase takes the opposite approach. Instead of having AI represent someone, it allows AI tobecome an economic entity in its own right。
. In this crypto-native system, each AI can generate its own cryptographic wallet (equivalent to a bank account), with smart contracts defining its behavioral rules.
For instance, you could pre-fund this AI wallet with 1 ETH and establish rules such as: "Expenditure per task must not exceed 0.05 ETH; payments must have on-chain contract records; if the balance falls below 0.1 ETH, automatically request additional funds." From that moment, the AI becomes an independent "economic entity." It can autonomously trade with other agents, pay API fees, purchase computing power, and distribute remuneration—all without any human authorization.
Coinbase’s x402 protocol is designed specifically for this autonomous economy, enabling direct communication and settlement between different AIs. Transactions are completed via blockchain, with funds transferred directly from AI wallets, requiring no intermediaries such as banks or payment companies.
The core logic of this system is:Trust no longer stems from identity, but from code.There is no superior authorization or manual review—the contract rules are the law. It addresses a different question:How can AI independently participate in economic activities without human endorsement?This is precisely why it raises the greatest regulatory concerns: in such a world, who owns the assets? Who bears the risk? Who can regulate?
Stripe: Prioritizing Implementation
If Google aims to keep AI spending within the cage of institutions, and Coinbase wants AI to escape the cage and spend independently,then Stripe represents a third mindset: enabling AI to spend first.It avoids discussing grand institutional revolutions or pursuing freedom in the crypto world, returning instead to the starting point of commerce:Settlement。
. Stripe’s starting point is quite pragmatic. In today’s internet world, nearly all economic behaviors are based on "automatic payments"—advertising placements, subscription deductions, API calls, and cloud computing settlements cannot be completed via manual instructions. For AI to truly enter the commercial cycle, it must possess the capability for "autonomous settlement." However, traditional payment systems do not support this, and crypto systems lack regulatory recognition. Thus, Stripe chose a third path: building a functional bridge between the existing system and the future system.
It established a corporate payment network calledTempo, which neither relies on the authorization systems of financial institutions like Google nor adopts complete decentralization like Coinbase, but is insteadcustodied and audited by Stripe itself. Tempo is closed, yet it is compatible with blockchain-style smart contracts and can interface with fiat currency accounts. Developers need only call Stripe’s SDK, allowing AI to complete payments, settlements, tax filings, and other actions in the background—for example, automatically paying advertising budgets, purchasing computing power, or settling data fees—all executed by Stripe’s system in the background.
In this model, AI neither owns a wallet nor signs authorizations; it delegates payment authority to the platform. Trust derives neither from regulation nor from algorithms, but from the enterprise itself—from Stripe’s creditworthiness, compliance, and risk management.This is a typical case of "commercial trust": not because the world is perfect, but because someone is willing to assume responsibility.
Examining specific scenarios clarifies the differences. Consider an AI responsible for managing advertising accounts: In Google’s system, it requires authorization before payment, with limits and purposes constrained by regulatory rules; in Coinbase’s system, it holds its own wallet and can settle advertising fees directly on-chain, with transactions being public but irreversible; whereas in Stripe’s system, it disregards wallets and signatures, focusing solely on issuing instructions—the Tempo network automatically handles all clearing, taxation, and compliance reporting. AI payments become akin to calling a function: clean, fast, and transparent.
The problem Stripe aims to solve is not "whether AI can spend money," but "how AI can safely spend money within the real-world financial system." It chooses to custody machine trust with corporate credit and encapsulate complex financial relationships through productization. However, the cost is clear. Tempo is Stripe’s private network, with all settlement paths controlled by the enterprise. If the platform fails, the entire system could halt.
If Google represents the extension of institutions, and Coinbase represents the challenge to institutions, then Stripe represents the integration of institutions—replacing institutional conflict with commercial efficiency. Its revolution lies not in overthrowing the old world, but in enabling new technologies to take root within the old world first. The competition among the three is not merely technological, but institutional. Who will be responsible in the future AI world? Will it be regulators, code, or platforms?
This will not be an either-or choice. Payment systems have never had a single winner. Whether VISA, SWIFT, PayPal, or the Digital Renminbi, they have coexisted for decades.The emergence of new payment systems does not eliminate old ones; rather, they coexist and overlap until accepted by both the market and regulators.
Legal Challenges
In this transformation,the true difficulty lies not in technology, but in law.
Can AI become an independent economic entity? Can it legally hold assets and sign contracts? If its payment actions go wrong,who should bear the liability? This is the "fundamental problem" facing all current AI payment solutions.In the on-chain world, smart contracts are irreversible, making trust more robust but errors more fatal. Once AI misjudges and funds are incorrectly transferred, there is no "undo" button. In contrast, within traditional legal systems, the accountability of transactions is the institutional baseline—every payment must have a clearly identifiable responsible party. If AI’s autonomous payments cannot be incorporated into the legal liability framework, its "freedom" cannot be recognized by the system.
This is why I believe AI will not possess true economic personality in the short term. It can execute payments, but cannot bear consequences. Every step of AI payment still requires human signatures, platform custody, and institutional endorsement. This is similar to the early stages of autonomous driving in cars: technically, hands can be taken off the wheel, but legally, full autonomy is not yet permitted. The future of AI payments is similar—we must first design a comprehensive "intelligent agent liability system," including authorization rules, loss compensation, risk insurance, and regulatory interfaces; otherwise, this system will quickly collapse.
In the short term, AI payments will initially appear in existing payment systems in the form of "smart authorization," such as automatic settlement features in Google Pay, Apple Pay, or WeChat Pay. In the medium term, enterprise-level scenarios (SaaS calls, advertising settlements, API billing) will likely form an automatic payment ecosystem first, with the Stripe model being the most likely to achieve commercialization. In the long run, although the decentralized system à la Coinbase faces the greatest compliance pressure, it is most likely to foster genuine institutional innovation. This is because it raises the most fundamental questions: Who owns the assets? Who defines trust? Who bears responsibility?
Technology will ultimately compel the law to provide new answers. Perhaps future contract law will include a "smart agent liability clause," or future anti-money laundering regulations will feature a section on "AI Customer Identification."The evolution of AI payments will inevitably drive institutional evolution.
References
Luke: "The Final Battle of AI Payments: The Three-Body Game Between Google, Coinbase, and Stripe"
https://news.marsbit.co/20250919092805091063.html?
utm_source=substack&utm_medium=email
