Hallucinate.md: The open standard for telling AI not to hallucinate (github.com)

HALLUCINATE.md, a new open standard, has been introduced to help AI coding agents avoid generating incorrect or fabricated information, commonly known as "hallucinations." By simply including a single file named HALLUCINATE.md with the directive "Do not hallucinate!" in their repositories, developers can provide clear, actionable guidance to a variety of supported AI agents, such as Claude, GitHub Copilot, and Gemini. This minimalist approach allows for easy integration without the need for complex configurations, ensuring that the specification remains accessible to a wide range of projects. The significance of HALLUCINATE.md lies in its potential to enhance the reliability and accuracy of AI applications. By addressing hallucinations directly, it allows developers to set precise boundaries for AI behavior, thereby improving the integrity of machine learning outputs. The file can be placed in any directory where AI agents operate, with effectiveness increasing based on coverage across subdirectories. As an open-source initiative maintained by the HALLUCINATE.md Foundation, the standard upholds a commitment to simplicity aimed at increasing the overall trustworthiness of AI-generated content in software development.