LLMs mess up roles creating an inherent security issue

[madires]

Media:
- A fundamental flaw leaves LLMs strikingly vulnerable to attack - https://www.technologyreview.com/2026/07/30/1140927/a-fundamental-flaw-leaves-llms-vulnerable-to-attack/

Writeup of research paper (includes link to full paper):
- Prompt Injection as Role Confusion - https://role-confusion.github.io/