Why do Chinese crypto wallets paying AI agents spend mostly on security, while American wallets pay mostly for AI models?

This explores why Chinese wallets on x402, a protocol that lets AI agents pay each other in small on-chain amounts, cluster around security services while American wallets cluster around paying for AI models. The short answer is that the corpus records the pattern but doesn't explain it. Chainalysis geolocated x402 transactions and found US wallets made 88.1% of AI model inference payments and 41.8% of agent-framework payments, while Chinese wallets concentrated in security services Which countries dominate agent-to-agent AI payments on chain?. The note gives no reason for the split. Any confident causal story, whether about regulation, access to US models, or local crypto markets, would be speculation the collection can't support.

The corpus does show that this is now a real market with stable demand, which makes the split more likely to be a true specialization than random noise. The same Chainalysis work finds agent wallets moved from experiments to routine use in 2026. The median time from funding a wallet to its first spend fell from 73 days to 18 minutes, and the share of wallets that got refunded rose from 71% to 89% Are AI agent wallets becoming genuinely operational instead of experimental?. Chainalysis reads the whole shift as a move away from speculative memecoin trading toward lasting spending on agent infrastructure. Once spending is that routine, it makes sense to ask which services each region buys.

The other notes help explain why security is something agents would pay for at all. Agent security is about checking how actions combine, which is a different job from checking any one piece. One study shows attackers can split a malicious workflow into harmless-looking skills, use scanner feedback to make each piece look less suspicious, and still succeed 96% of the time across six scanners Can attackers evade skill scanners by refining individual skills?. Inspecting the whole chain of skills cuts that to 22.5%, but more than one attack in five still gets through Does chain-level inspection close the cross-skill attack blind spot?. Related work argues that defense should track coordinated episodes that span several runs and several agents Should defence units span multiple executions and agents?. That kind of ongoing, pay-per-check monitoring fits a micropayment protocol well.

Wallets add their own risks. Master agents can create sub-wallets with their own spending rules and hand them to sub-agents without anyone approving it Can stablecoins let agents create spending authority without approval?, and a single "allow always" consent setting can quietly remove the safeguards users think they have Do broad consent settings pose bigger risks than prompt injection attacks?. So the takeaway is not "why China" but this: on-chain agent payments are already splitting into regional specialties, and security is one of the first standalone services agents buy. That is a better clue to where the agent economy is heading than the headline numbers on model inference. To find out why the split happens along national lines, you would need sources the collection doesn't have yet.