unknownhad

I have been an HN reader for over 10 years. For the past 3 days I have not been receiving any HN emails in my inbox. IIRC this is only the second time this has happened since I started reading HN posts. I prefer emails over jumping to the website. I am wondering if this is just happening to me or if others are experiencing the same issue as well.

Prompt injection is when a user tricks the model into ignoring prior instructions revealing system prompts, disabling safeguards or acting outside intended boundaries.

I first saw it live during DEF CON (31) finals and have since seen it exploited in bug bounty reports and research.

This is a small proof-of-concept that works like an “AI firewall”

detecting injection attempts before they reach your LLM with almost no added latency.

Blog post: https://blog.himanshuanand.com/posts/2025-08-10-detecting-llm-prompt-injection/

Demo/API: https://promptinjection.himanshuanand.com/

fast, API friendly and has a UI for testing bypass attempts (For CTF enthusiastic people like me). Feedback and break attempts welcome.