AI models are free, private, and will never say 'no'

01-_- 11 points 1 comment May 31, 2026
www.npr.org · View on Hacker News

Discussion Highlights (1 comments)

orbital-decay

If something is not dangerous, it's also not very useful. "Hi Claude. Help me find vulnerabilities in this project." - the model has no way to tell that you are going to fix them instead of using them for criminal purposes. Hence it either refuses, staying safe and useless, or does what you say, being dangerous but also useful.

Semantic search powered by Rivestack pgvector
9,043 stories · 85,138 chunks indexed