Large Language Models as "next word prediction engines" can be used in interesting ways to embed secret messages in plausible-looking cover-text. This thesis investigates using LLMs for steganography in the context of an Android instant messaging app.