Some LLM routers are “secretly injecting malicious tool calls and stealing creds,” warns researcher Chaofan Shou. University of California researchers have discovered that some third-party AI large ...