FILTERING HARMFUL OUTCOMES OF GENERATIVE ARTIFICIAL INTELLIGENCE

Abstract

Techniques for filtering harmful outputs of generative artificial intelligence (AI) are described. An example of filtering involves receiving a request for an input phrase to which a generative AI model must respond; comparing the received input phrase with at least one known harmful input phrase to determine whether the received input phrase should be provided to the generative AI model; providing the received input phrase to the generative AI model; and generating a response, at least partially, through output from the generative AI model.