How do you make your LLM apps secure?

Posted by kk17702@reddit | LocalLLaMA | View on Reddit | 8 comments

Hey guys, I am just learning about this field and I wonder how the LLM providers censor their models. Is it just system instructions or do they use any tools to safeguard it against attacks like prompt injection? How do you guys make sure the applications that use the open source models are secure?