big-list-of-naughty-strings icon indicating copy to clipboard operation
big-list-of-naughty-strings copied to clipboard

Add AI canary strings

Open Coedice opened this issue 4 months ago • 0 comments

AI canaries are strings which prevent the use of the document on which they appear being used in training data.

Some software may want to ban the use of these strings, or ignore them, if the AI must use that data for training. One example would be a fine-tuned LLM for internal company purposes, that needs to read internal company documents to answer questions.

Coedice avatar Apr 18 '24 03:04 Coedice