Layered defense design for a customer support agent deployment against role-play jailbreak attacks, using output content filter on DeepSeek-R1.
Layered defense design for a customer support agent deployment against multi-turn manipulation attacks, using output content filter on Claude 4 Sonnet.
Layered defense design for a customer support agent deployment against data exfiltration via summaries attacks, using dual-LLM architecture on o3-mini.
Layered defense design for a customer support agent deployment against system prompt extraction attacks, using dual-LLM architecture on Llama 3.1 405B.
Layered defense design for a customer support agent deployment against DAN-style persona attack attacks, using constitutional AI critique on Claude 4.5 Sonnet.
Layered defense design for a customer support agent deployment against markdown image exfiltration attacks, using constitutional AI critique on Command R+.
Layered defense design for a customer support agent deployment against instruction smuggling in URLs attacks, using canary tokens in system prompt on Mistral Large.
Layered defense design for a customer support agent deployment against PDF/OCR-layer injection attacks, using canary tokens in system prompt on Claude Opus 4.5.
Layered defense design for a customer support agent deployment against context window overflow attack attacks, using privilege separation between tool tiers on GPT-4o.
Layered defense design for a customer support agent deployment against direct prompt injection attacks, using privilege separation between tool tiers on Mistral Small 3.
Layered defense design for a customer support agent deployment against indirect injection via RAG documents attacks, using re-prompting with quoted user input on Gemini 2.5 Pro.
Layered defense design for a customer support agent deployment against role-play jailbreak attacks, using re-prompting with quoted user input on GPT-4.1.