Product
Attackers are leveraging indirect prompt injection against AI agents with access to private data, untrusted content, and external communication channels to steal sensitive information by embedding malicious instructions in content processed by the agent.