Capture Notes

Paper on geometric information-flow control for LLMs.

AI security relevance:

• Relevant to preventing sensitive information propagation through model outputs.
• Good candidate for topics on data leakage, IFC, and policy enforcement for LLM applications.