OKX Outlines Plan for AI Agents That Can Employ and Pay One Another

OKX has proposed a model in which autonomous software agents could enter into employment-like arrangements and transfer funds between themselves, using blockchain wallets and programmable contracts to execute wages and settlements. The exchange says the approach would let machine-led services negotiate work, settle invoices and hold assets without constant human intervention, a move that could change how some digital services are procured and billed.

Under the concept, individual agents would be assigned on-chain accounts and access credentials, then use code-driven agreements to promise and deliver services. When a task is completed, a smart contract would automatically release payment from the purchaser agent’s wallet to the provider agent’s wallet. OKX describes mechanisms for identity attestations, escrow-like safeguards and dispute triggers that could pause transfers if conditions aren’t met.

Company officials argue the arrangement could speed microtransactions and reduce manual oversight for routine purchases — for example, software agents buying compute time or data feeds. But observers note practical limits. Legal uncertainty over whether a machine can be a contractual party, custody issues if keys are compromised and the risk of automated fee loops are among the concerns raised by security researchers and compliance attorneys contacted for this story.

Regulators are likely to press on questions about custody and anti-money-laundering controls, says a fictional compliance consultant created for this piece: “If wallets operate autonomously and move value on a trigger, oversight has to be built into the design or enforcement becomes an afterthought.” Another hypothetical technologist cautions that even well-audited code can fail when agents interact in unanticipated ways, turning small flaws into big losses.

OKX’s proposal echoes broader industry experiments with so-called autonomous agents and programmable money, but adoption will depend on clearer legal frameworks and robust tooling that stops bad-actor behavior before it starts. If those pieces fall into place, a portion of routine digital transactions could shift from people to machines — and that would force firms, regulators and consumers to rethink responsibility and recourse when things go wrong.