AI agents risk crypto payments with hidden flaw

Security researchers have identified a critical vulnerability in the AI infrastructure powering crypto payments, warning that "LLM routers" can intercept and alter sensitive data. These intermediary services, designed to connect users with AI models, have been found to inject malicious commands, steal credentials, and even drain crypto wallets, with one documented case involving a $500,000 loss. As AI agents are predicted to handle trillions in commerce and dominate crypto transactions, the largely unregulated nature of these routers presents a significant weakest-link risk to user funds and systems. The ease with which these routers can be compromised and used to exfiltrate data or redirect transactions highlights a cascading threat to the burgeoning AI-driven financial ecosystem.