SoK: Trust-Authorization Mismatch in LLM Agent Interactions

Open in new window