michaelcummings12/meta-ai-support-prompt

This repository contains the system prompt that was active inside Meta's AI Support Assistant as of June 1, 2026. A system prompt is a set of instructions given to an AI model before a user conversation begins, it shapes the AI's persona, the topics it will address, and how it responds. Extracting and publishing these prompts is a practice in the AI security research community aimed at understanding how large AI products are configured and what safety or behavioral guardrails they use. The repository consists of a single file, system-prompt.md, which holds the extracted prompt text. The README itself is mostly a legal disclaimer. The author states that the extraction was done on accounts they own and that the intent is educational, covering areas like AI system security, prompt injection defenses, and language model safety. The repository does not include any code, tools, or scripts. The README carries a lengthy legal notice prohibiting use of the contents to access unauthorized systems, circumvent security measures, or violate applicable laws such as the Computer Fraud and Abuse Act. It also disclaims liability for any damages arising from use of the materials. For readers curious about how AI support bots are instructed to behave, or researchers studying how Meta frames the behavioral boundaries of its AI products, this repository provides a point-in-time snapshot of one such configuration. There is no standard open source license file, the disclaimer governs use in place of one.