The Rise of AI Agents and the Battle to Prove They’re Human-Backed

A Trillion-Dollar Question: Who’s Behind the Bot?

Imagine a future where artificial intelligence handles your shopping, books your restaurant reservations, and manages your online errands while you focus on more important things. It sounds convenient, right? But here’s the catch: as AI agents become more prevalent in conducting business online, we’re facing a fundamental challenge that threatens to reshape how the internet works. Industry experts predict that by 2030, AI-driven commerce could balloon into a market worth anywhere from $3 trillion to $5 trillion. That’s not pocket change—it’s a complete transformation of how transactions happen online. Yet with this massive shift comes a critical question: how do we ensure there’s actually a real human being authorizing all this autonomous activity? This isn’t just a technical problem; it’s a question that gets to the heart of trust, accountability, and the future of digital commerce. Enter World, the identity verification project backed by OpenAI’s Sam Altman, which believes it has cracked the code on proving humanity in an increasingly automated world.

World’s Solution: AgentKit and the Human Verification Challenge

World, the company formerly known as WorldCoin, has unveiled its answer to the AI identity crisis with a new toolkit called AgentKit. Released this week, this developer-friendly system allows AI agents to carry cryptographic proof that they’re authorized by a genuine human being through World’s existing World ID verification system. Think of it as a digital certificate of authenticity—except instead of verifying a painting or a signature, it’s confirming that somewhere behind that AI agent making purchases or booking services, there’s an actual person who gave it permission to act. The technology works in tandem with x402, an innovative protocol developed by Coinbase and Cloudflare that facilitates what insiders call “agentic payments.” This system embeds stablecoin micropayments directly into the internet’s communication infrastructure, enabling AI agents and software programs to pay each other without any human having to manually approve each transaction. As Erik Reppel, who heads engineering at Coinbase Developer Platform and founded x402, puts it succinctly: “Payments are the ‘how’ of agentic commerce, but identity is the ‘who.'” His point is crucial—knowing how money moves is one thing, but understanding who authorized that movement is what transforms AI agents from potential security threats into legitimate participants in the digital economy.

The Explosive Growth of Agentic Commerce

The timing of World’s announcement couldn’t be more relevant. AI agents are evolving at breakneck speed, taking on increasingly complex tasks that most of us find tedious or time-consuming. From scanning multiple e-commerce sites to find the best price on that gadget you want, to negotiating the maze of online booking systems to secure that perfect dinner reservation, AI agents are becoming our digital personal assistants. The numbers tell a compelling story about where this is headed. Some forecasts suggest that agentic commerce could capture up to 25% of all U.S. e-commerce by 2030, contributing to that massive $3-5 trillion market projection. Industry heavyweights are taking these predictions seriously. Brian Armstrong, the founder of Coinbase, has stated he believes we’ll “very soon” reach a point where AI agents outnumber humans in conducting transactions online. Changpeng Zhao, who founded Binance, went even further with his prediction, suggesting that AI agents will eventually conduct one million times more payments than people do—and he’s convinced they’ll use cryptocurrency to do it. These aren’t fringe voices; these are leaders of major financial technology companies betting big that autonomous AI commerce is the future.

The Identity Gap That Payments Can’t Fill

Here’s where things get tricky. While the technology for AI agents to make payments is advancing rapidly, there’s a fundamental problem that payment systems alone can’t solve: proving identity and preventing abuse. DC Builder, a research engineer at the World Foundation, highlighted the core issue with a simple example: “One person could run thousands of agents that all pay small fees.” Without some way to verify that agents are tied to real humans, nothing stops someone from deploying an army of bots to exploit free trials, overwhelm systems with fraudulent bookings, or game platforms that offer promotional benefits to new users. World’s AgentKit addresses this vulnerability by creating a verifiable link between multiple AI agents and the single human who controls them. This connection happens at the identity level, allowing platforms to enforce meaningful limits. A World spokesperson explained the practical applications: “AgentKit allows developers to link multiple agents to the same verified human. This means a platform can allow someone to run several agents while still enforcing limits based on the underlying person.” In practice, this could mean a service could offer one free trial per person—not per agent—or limit bookings to a certain number per day per human, regardless of how many AI assistants that person has deployed. This prevents exploitation while still allowing legitimate users the flexibility to employ multiple specialized agents for different tasks.

Solving the Bot Detection Problem

There’s another significant hurdle that agentic commerce faces: most websites today are designed to treat automated traffic with suspicion, and many actively block bots altogether. This defensive posture makes sense when you consider the internet’s long history of malicious bots scraping content, creating fake accounts, and attempting fraud. However, this blanket approach to bot-blocking is increasingly incompatible with a future where legitimate AI agents routinely act on behalf of real users. AgentKit offers a potential solution by allowing users to delegate their World ID—a privacy-preserving proof of their unique humanity—to AI agents acting under their authority. Importantly, World isn’t positioning this as a winner-take-all solution that must replace every other identity verification system. A company spokesperson told CoinDesk: “This isn’t necessarily an either-or choice. World ID is designed to be a proof of human layer that developers can use on their own or alongside other identity systems.” The technical implementation uses zero-knowledge proofs, a cryptographic method that allows platforms to verify an agent represents a real person without actually collecting or storing any personal information about that individual. According to World, this privacy-preserving design isn’t just a nice feature—it’s an essential requirement for scaling identity verification in an internet increasingly dominated by AI activity.

Beyond the Orb: Future Verification Methods and Market Impact

Currently, AgentKit in its beta version relies on Orb-based biometric verification—World’s most recognizable and controversial component, involving specialized hardware that scans users’ irises to confirm their unique human identity. However, the company has announced plans to expand beyond Orb verification to include additional credential types. Future iterations will support NFC-enabled passports and government-issued IDs through “World ID Credentials,” giving users alternative methods to prove attributes about themselves without revealing sensitive personal information. At the time of the announcement, World’s real-time verification counter showed over 17.9 million humans had been verified through their system, making their network among the largest proof-of-personhood systems operating globally. This existing user base positions World to potentially become the default identity layer for an internet increasingly populated by AI agents acting on human behalf. The broader implications extend far beyond just one company’s technology. As Visa and Coinbase both develop their own approaches to facilitating agentic commerce, we’re witnessing the early stages of what could be fundamentally different visions for how the internet operates. The question isn’t whether AI agents will become commonplace in online commerce—most experts agree that’s inevitable. The real question is which identity verification systems will earn the trust of both platforms and users, and whether the internet of the future will prioritize privacy, convenience, security, or some carefully balanced combination of all three. As this technology moves from beta testing to mainstream deployment, the decisions made today about identity verification will shape how billions of people and their AI assistants interact with the digital economy for decades to come.