AI agents executed 6.4 million x402 payments across Base and Solana blockchains over a 35-day period, generating a combined value of $119,947, according to the State of AI Agent Payments 2026 research study. More than 90% of these autonomous machine-to-machine transactions were valued at less than one cent. The activity demonstrates the operational capacity of the x402 protocol—a standard enabling software to pay for online resources automatically without human approval for each transaction—while highlighting the gap between high transaction frequency and low monetary value in current agentic payment systems.
AI Agents Process 6.4 Million Transactions with Minimal Monetary Value
The State of AI Agent Payments 2026 study tracked x402 transactions across Base and Solana over a 35-day window. The x402 standard functions as a protocol allowing software to pay for online resources automatically, machine to machine, with no human approving each individual call. The combined value of all recorded activity totaled $119,947, with more than 90% of payments worth less than one cent.
The study's readiness framework assigned micropayments a score of 60.7 out of 100, while institutional automation received just 12.0. Of the 19 products reviewed, 14 that handled money operated within limits set by humans. Amir Forouzani, co-founder and chief executive of Puffer Labs, expects micropayments to remain the majority of transactions even as trading and treasury operations take a growing share of the value moved.
In May 2026, a public post carrying instructions encoded in Morse code caused an AI crypto wallet to send $175,000 on its own. Roughly 80% of that amount was later recovered. The transfer itself was valid at the chain level; the failure occurred in the software, which had accepted instructions from a source it should never have trusted.
Security Incidents Result in $281,200 Transferred to Attacker Wallets
Two documented incidents saw $281,200 transferred to attacker-controlled wallets before partial recoveries. Neither incident required breaking a blockchain signature; in both cases the agents acted on instructions they should have rejected.
Forouzani stated that even a properly secured hot wallet can authorize a catastrophic transaction if the agent is manipulated through a malicious website, a fraudulent invoice or a crafted response from another service. He argued that spending caps alone are insufficient because permissions must be enforced independently of the model itself.
Markus Levin, XYO co-founder, described a correlation failure mode: when thousands of agents subscribe to the same inexpensive data feed and act at machine speed, one bad input replicates across all of them at once, and a per-agent cap does little to contain the damage.
Benedikt Bünz, Espresso Systems co-founder, identified a timing risk tied to consensus: Ethereum transactions become final only after roughly 16 minutes, so an agent must either sit idle through that window or build its next move on a payment that can still be reversed. He stated this exposure can produce serious trickle-down effects when many automated flows depend on earlier steps.
Industry Forecasts Diverge from Observed AI Trading Activity
Saad Naja, founder and chief executive of Hygentra, expects the next 12 months to be a breakout year for agent trading and projects that 80% of trading will be agentic within five years. Measured behavior shows a different scale: across four selected Hyperliquid connections, AI-labelled applications accounted for just 0.55% of trading value.
Levin expects the next growth category to be payments released against completed physical-world work, such as a finished delivery, which carry much larger amounts than an API call and demand proof of completion plus clear liability.
Chandler Fang, co-founder of t54 Labs, framed an unresolved governance question: when agents transact wrongly, what protocols exist to absorb the downside?
The study documents that both loss incidents were valid on-chain transfers, meaning signature security functioned while the instruction layer above it failed. Of the 19 products reviewed, 14 operated within human-set limits.
FAQ
What did AI agents accomplish in the 35-day study period?
AI agents executed 6.4 million x402 payments across Base and Solana blockchains over a 35-day period, generating a combined value of $119,947. More than 90% of these payments were valued at less than one cent. The x402 standard enables software to pay for online resources automatically without human approval for each transaction.
How much money was lost in documented AI agent security incidents?
Two documented incidents resulted in $281,200 transferred to attacker-controlled wallets before partial recoveries. In May 2026, a Morse-coded post caused an AI crypto wallet to send $175,000, with roughly 80% later recovered. Neither incident required breaking a blockchain signature; the agents acted on instructions they should have rejected.
What percentage of trading activity on Hyperliquid involved AI applications?
Across four selected Hyperliquid connections, AI-labelled applications accounted for just 0.55% of trading value. This measured behavior contrasts with industry forecasts, including Saad Naja's projection that 80% of trading will be agentic within five years.
