
Blockchain intelligence firm finds that most activity on the x402 payment protocol stems from scripts rather than autonomous AI agents.
AI-generated summary
Coinbase launched the x402 protocol in 2025 to integrate payments directly into web requests. The protocol allows facilitators to verify payment authorizations and submit blockchain transactions on behalf of users.
Most payment volume on Coinbase’s x402 protocol isn’t coming from AI agents, according to blockchain intelligence company TRM Labs.
TRM’s report, published Wednesday, examined $52.7 million across 198.9 million settlements processed by known x402 facilitators on Base, Solana, and Polygon since May 2025. After excluding self-payments and other anomalous flows, researchers estimated that just 0.6%–7.5% of the remaining commerce by value appeared to come from AI agents.
“The assumption is that a true agent explores across multiple services and products, while an address repeating the same price behaves more like a script hitting one service over and over,” the report said. “This is a deliberate modeling choice, and it may understate the space: many agents today could be single-purpose, paying one service repeatedly, which this test would read as a script.”
Launched by Coinbase in 2025, x402 integrates payments into web requests. The buyer receives a price and signs a payment authorization; a facilitator verifies it, submits the blockchain transaction, and pays the network fee.
According to TRM Labs, an ordinary script can follow that sequence without an AI agent. Scheduled jobs, load tests, and self-dealing can also produce the same blockchain records, making protocol totals insufficient to measure agent commerce.
TRM removed self-payments, bulk flows from one or two payers, and sellers with fewer than 10 buyers, leaving $25.62 million in likely commerce. It then screened for facilitator-broadcast payments with varying amounts averaging under $1. A stricter test required that pattern across months, plus public agent registration or payments to multiple sellers. TRM cautioned that these criteria could miss genuine agents repeatedly buying the same service.
In late 2025, activity included apparent meme-token minting and payments to one AI-analysis service. Volume concentrated in a single payment contract in early 2026, before AI-service payments returned through an agent-payment router around midyear. Across the full period, USDC accounted for 99.6% of settled value, or $52.47 million.
Companies have continued investing in agent payments. Binance’s August Agent OS launch included an x402 payment layer, while Coinbase’s Base targeted agent and payments startups through its $1 million accelerator. Amazon announced AgentCore Payments with Coinbase and Stripe in May, enabling agents to pay for online services with stablecoins.
Beyond measuring usage, TRM identified gaps in assigning responsibility for agent payments.
“On-chain agent registries let individuals declare ownership of an agent address,” TRM Labs wrote. “However, this declaration is voluntary and currently not utilized by the majority of participants.”
TRM called for more accurate registration, counterparty reputation information that agents can check, and monitoring suited to large numbers of small payments.
“The rail already works. What is needed is accurate registration, counterparty reputation an agent can check on its own, and monitoring built for volume rather than value,” they wrote. “Agentic commerce will need agentic compliance.”

Under the EU's Cyber Resilience Act, manufacturers of connected hardware wallets and software must report actively exploited vulnerabilities or severe security incidents to cyber authorities within 24 hours, effective September 11, 2026.

Anthropic CEO Dario Amodei warned that rapid AI development risks outrunning human control through recursive self-improvement, proposing international safety standards supported by OpenAI CEO Sam Altman.

Fintech company Revolut exposed sensitive customer data including IDs and transaction histories after receiving a fraudulent information request from a spoofed government email domain.

The XRP Ledger has activated the fixCleanup3_3_0 amendment, introducing maintenance updates for transaction handling and AMMs. Servers failing to support the new rules risk becoming amendment-blocked, necessitating software updates for infrastructure operators.

OpenAI's new ChatGPT Images 2.5 model is compared against Google's Nano Banana 2. While OpenAI resolved previous oversharpening issues, the models remain closely matched in performance, with each showing specific strengths in text rendering, spatial awareness, and factual accuracy.

Greenberg Traurig reported an unauthorized actor accessed and posted limited documents on the dark web, exposing client Social Security information. The breach is part of a growing trend targeting law firms, with BakerHostetler handling nearly 60 cybersecurity incidents in 2025. Other firms including Taft Stettinius, Herbert Smith Freehills, WilmerHale, Goodwin Procter, and Quinn Emanuel disclosed breaches. Crypto companies Coinbase, Ledger, and SafePal also reported customer data exposures.