Bookmark · 25 June 2026, 13:28

Prompt Injection as Role Confusion

• ml-security
• prompts