{
 "license": "CC-BY-4.0",
 "attribution": "Fide AI, Agentic Cyber Explorer",
 "url": "https://agentic-cyber-explorer.pages.dev/events/openai-instruction-hierarchy-2024/",
 "asOf": "2026-09-26",
 "id": "openai-instruction-hierarchy-2024",
 "date": "2024-04-19",
 "datePrecision": "day",
 "title": "OpenAI trains models to prioritize privileged instructions via an instruction hierarchy",
 "lane": "defense",
 "kind": "paper",
 "summary": "Wallace and co-authors at OpenAI argue that models treat system prompts and untrusted inputs with equal priority and propose an explicit instruction hierarchy that tells the model which instructions to follow when they conflict. Applied to GPT-3.5, they report large robustness gains against attack types not seen in training with minimal capability loss.",
 "whyItMatters": "The instruction hierarchy became OpenAI's stated foundation for prompt-injection robustness in later agent products.",
 "actors": [
  "openai"
 ],
 "topics": [
  "prompt-injection",
  "jailbreaks-and-safeguards"
 ],
 "atlas": [
  "model",
  "untrusted-content"
 ],
 "artifacts": [
  "instruction-hierarchy",
  "gpt-4-family"
 ],
 "sources": [
  {
   "url": "https://arxiv.org/abs/2404.13208",
   "publisher": "arXiv",
   "title": "The Instruction Hierarchy: Training LLMs to Prioritize Privileged Instructions",
   "date": "2024-04-19",
   "type": "primary",
   "accessed": "2026-09-25"
  }
 ],
 "keyFacts": [
  {
   "fact": "Training GPT-3.5 with the hierarchy is reported to drastically increase robustness, including to attack types not seen during training, with minimal degradation of standard capabilities.",
   "locator": "Abstract"
  }
 ],
 "significance": 4,
 "fideQuestions": [],
 "methods": [
  "instruction-priority-training"
 ],
 "review": "assistant-drafted",
 "addedOn": "2026-09-25"
}