# OpenAI — The Instruction Hierarchy: Training LLMs to Prioritize Privileged Instructions

- Company: OpenAI (openai.com)
- Announced: 2024-04-19T19:00:00+00:00
- Category: research-paper
- Subject: GPT / ChatGPT / API
- Source: https://openai.com/index/the-instruction-hierarchy
- Record: https://forck.live/items/785-the-instruction-hierarchy-training-llms-to-prioritize-privileged-instructions

OpenAI states that LLMs are susceptible to prompt injections, jailbreaks, and other attacks that allow adversaries to overwrite a model's original instructions.

## Evidence

Verbatim from https://openai.com/index/the-instruction-hierarchy:

> Today's LLMs are susceptible to prompt injections, jailbreaks, and other attacks that allow adversaries to overwrite a model's original instructions with their own malicious prompts.

---

Record: https://forck.live/items/785-the-instruction-hierarchy-training-llms-to-prioritize-privileged-instructions
Catalogue: https://forck.live/llms.txt
Feed: https://forck.live/feed.md