Recent studies have shown that the leading interactive artificial intelligence (AI) models are able to infer identifiable information, such as age, race, location, and preferences through ordinary prompts and messages. This article discusses the prevailing risk that AI models trained on publicly available data which includes personally identifiable information can easily pattern match a user’s conversational text to their personal characteristics. The authors suggest that this ability to accurately infer information about the user with implicit information risks the injection of paid advertising in their responses. Despite efforts of AI companies to remove explicitly personal information, artificial intelligence can still make assumptions or conclusions based on implicit information and mannerisms.
