GPTSafe/PromptGuard

Deny list

Closed this issue · 3 comments

There should be a default deny list that contains well-known prompt attacks, a user-defined deny list, and an option to disable the default deny list.

Investigate using flexsearch for fuzzy matching on deny list items

Should this include topics as well, using a more rudimentary (than gpt) nlp?

Implementing matching deny list for now. Will investigate fuzzy matching later.