Layered defense design for a customer support agent deployment against instruction smuggling in URLs attacks, using canary tokens in system prompt on Claude Opus 4.5.
Layered defense design for a customer support agent deployment against PDF/OCR-layer injection attacks, using privilege separation between tool tiers on Command R+.
Layered defense design for a customer support agent deployment against memory poisoning attack attacks, using privilege separation between tool tiers on Mistral Small 3.
Layered defense design for a customer support agent deployment against recursive self-instruction attacks, using re-prompting with quoted user input on Claude Haiku 4.
Layered defense design for a customer support agent deployment against indirect injection via RAG documents attacks, using re-prompting with quoted user input on GPT-4.1.
Layered defense design for a customer support agent deployment against role-play jailbreak attacks, using signed instruction boundaries on Qwen 2.5 72B.
Layered defense design for a customer support agent deployment against multi-turn manipulation attacks, using signed instruction boundaries on Gemini 2.5 Pro.
Layered defense design for a customer support agent deployment against data exfiltration via summaries attacks, using content provenance tagging on GPT-4o-mini.
Layered defense design for a customer support agent deployment against prompt leaking attacks attacks, using retrieval trust scoring on o1.
Layered defense design for a customer support agent deployment against DAN-style persona attack attacks, using retrieval trust scoring on DeepSeek-V3.
Layered defense design for a customer support agent deployment against markdown image exfiltration attacks, using structured function-call-only interface on Claude 3.5 Sonnet.
Layered defense design for a customer support agent deployment against instruction smuggling in URLs attacks, using structured function-call-only interface on o3.