According to Mark Cuban, the reliance on trade secrets over patents is set to increase due to the potential misuse of patents by large language models (LLMs). Cuban argues that once a patent is filed, LLMs can train on the disclosed information, enabling users to generate competitive alternatives and effectively undermining the original intellectual property. This shift has significant implications for businesses aiming to protect their innovations in the AI-driven landscape.