AI Agents Must Become Paying Customers of Human Knowledge

An OpenAI researcher messaged company president Greg Brockman about "a hack to get around nytimes paywall." Brockman replied, "ah nice." The exchange appears in newly unsealed court filings in The New York Times' copyright case against OpenAI and Microsoft.

The exchange is jarring. It also exposes a structural gap in the emerging AI economy: machines are becoming some of the largest consumers of human knowledge, and there is no working system for them to pay for it.

OpenAI and Microsoft argue their use of copyrighted material falls under fair use. Publishers disagree. Courts will decide. But the AI economy needs more than a judicial answer, argues a Fortune commentary by a payments-industry veteran who has worked as a financial regulator, an international standard setter and a C-level payments executive, and is now writing a book on AI agents authorized to act and pay on consumers' behalf.

The missing piece, the author argues, is an architecture that preserves three things the internet is pulling apart: rights, credit and compensation.

Micropayments, finally with a use case

One of the most practical elements of that architecture may be an old idea whose moment has arrived: micropayments powered by agentic payments.

An AI agent may need one paragraph from a newspaper, one data point from a research service, one photograph or one court opinion. It may need it once, for seconds. A monthly subscription is the wrong instrument. Free access is not a sustainable answer either.

Micropayments have existed for decades but saw little practical use. Transaction costs mattered, but the bigger obstacle was human: people do not want to make hundreds of tiny purchasing decisions a day, and the operational overhead rarely justified the revenue.

AI agents do not have that problem. They can make thousands of such decisions programmatically.

There is still considerable debate over whether agentic payments will become a major economic layer. The author, drawing on work on agentic payments and the trust infrastructure they require, concludes they will.

Narrow mandates, not blank checks

Agents will not only discover products. They will increasingly complete transactions under a consumer's instructions: buying a news article for a few cents, paying for computing power according to actual usage, or executing smart contracts that settle automatically once agreed conditions are met.

Agents will also transact with other agents. An algorithm acting for a buyer may evaluate price, timing and terms against another algorithm acting for a seller, completing the deal automatically within parameters set by the humans or businesses behind them.

This does not mean handing agents unlimited access to money. Delegation will more likely be narrow: defined mandates, spending limits, approved categories and explicit conditions. Buy this if it costs less than $20. Purchase this article if it costs less than 10 cents. Renew this service only if the price stays within a specified range.

Agentic payments can run across several financial rails: cards, account-to-account payments supported by open banking, bank transfers, stablecoins, prepaid balances and new machine-native protocols. For some cross-border micropayments, stablecoins and other programmable rails may make very small transactions faster and cheaper than traditional methods.

The important development is not one particular rail. It is the creation of an ecosystem in which software can identify a transaction, verify authority, select an appropriate payment method and execute it within defined parameters.

That makes previously impractical models commercially viable: paying fractions of a cent for data, a few cents for an article, charging by token or API call, or automatically splitting revenue among multiple rights holders.

The trust layer

Moving money is only part of the challenge. Counterparties need to know which agent they are dealing with, who authorized it, the scope of its mandate, the user's intent and its spending limits. "Know your agent," verifiable intent, permissioning, fraud prevention, audit trails and dispute mechanisms are becoming part of a new trust architecture for delegated machine action.

Those same capabilities fit the copyright problem remarkably well.

Copyright is not merely a payment claim. It recognizes that creative work has an author, and that principle should not disappear because the reader is now a machine.

The author's model: content carries machine-readable information identifying its creator, rights holder, permitted uses, attribution requirements and price. An agent requests access and checks those terms against its mandate. If the use is permitted, it pays automatically and carries the provenance forward. If attribution is required, it stays attached. If the use is prohibited, payment does not make it permissible. Smart contracts could distribute proceeds automatically among relevant rights holders.

The building blocks already exist

The infrastructure is appearing. Mastercard's Agent Pay for Machines is designed for continuous machine-to-machine transactions, including microtransactions worth fractions of a cent. Visa's Trusted Agent Protocol focuses on verifying agent identity and authorization. Stripe and Tempo's Machine Payments Protocol provides an open standard for programmatic, machine-native transactions.

None of this will resolve disputes over historical training data, nor should it replace copyright law. It would give copyright something it has never had at internet scale: an operational layer built for machines.

The stakes go beyond publishers' revenues. If AI systems increasingly substitute for visits to original sources while returning little value or recognition to creators, the incentive to produce expensive, original human knowledge weakens. Meanwhile, the web is filling with synthetic material, and future models increasingly learn from earlier ones.

The answer cannot be to lock knowledge behind walls and hope increasingly capable machines never cross them. Nor should it be to declare human knowledge free raw material for AI.

Copyright can establish the rights. Provenance can preserve authorship and credit. Agentic payments can make compensation granular, automatic and economically viable at machine scale.

If AI agents are going to become some of the world's largest consumers of human knowledge, they should also become paying customers — recognizing who created what they use, respecting the rights attached to it, and paying for the value they consume.