Security of AI Agents

Capture Summary

Paper discussing security issues of AI agents and defense methodologies at the component level of an AI agent architecture.

Relevance

• Useful as an architectural baseline for decomposing agent attack surfaces.
• Helps distinguish model-level safety failures from agent component security failures.
• Candidate for source-to-evidence extraction around component-level defenses.

Collection Notes

Collected as foundational agent security architecture research.