
What Is Indirect Prompt Injection and How Does It Reach AI Agents?
An instruction a third party plants in content an agent reads, such as an email, file, web page or tool result, so the attacker never touches the prompt.
Research topic
Research on delivery paths: AI agent security, tokenization and checks for sensitive data access before allowing enterprise workflows to run.

An instruction a third party plants in content an agent reads, such as an email, file, web page or tool result, so the attacker never touches the prompt.