Hacker News new | ask | show | jobs
by zwnow 295 days ago
Doesn't help.

https://gandalf.lakera.ai/baseline

This thing models exactly these scenarios and asks you to break it, its still pretty easy. LLMs are not safe.