The Dual LLM Pattern for Building AI Assistants That Can Resist Prompt Injection

Metadata

Author: Simon Willison’s Weblog
Full Title: The Dual LLM Pattern for Building AI Assistants That Can Resist Prompt Injection
URL: #atom-everything

I really want an AI assistant: a Large Language Model powered chatbot that can answer questions and perform actions for me based on access to my private data and tools (View Highlight)
So, if it turns out we can’t solve this class of vulnerabilities against the design of existing Large Language Models, what’s a safe subset of the AI assistant that we can responsibly build today? (View Highlight)