An important distinction in Plural's application of AI is in the type of data Plural is ingesting to apply AI features.
All of the data that Plural ingests is publicly available bill text. It's the language of the bills and policies being proposed.
Plural does not use any PII (Personally Identifiable Information) in training its AI models.
Because Plural's models are trained on and operate on bill data, and only bill data, there is also less room for bias or outside noise. This is in contrast to training an AI model that sources information on legislation from news written about that legislation -- an approach that can bring in the potential for bias and also leaves plenty of bills (those that haven't been covered in the news) without the data needed to accurately supply the model.
We at Plural acknowledge that there is risk in using AI for public policy. This is why we also supply all of the information you need to gather insights yourself and fact-check AI insights alongside our AI features. We also encourage you to consistently check AI-powered work and let us know of inaccuracies so that we can continuously strengthen our models.