{
 "id": "AIR-2020-0024",
 "slug": "gpt-2-model-replication-by-external-researchers",
 "title": "GPT-2 Model Replication by External Researchers",
 "summary": "Researchers successfully replicated OpenAI's GPT-2 language model prior to its full release using publicly available technical information. The event highlighted risks related to tiered release schedules.",
 "type": "attack_on_ai",
 "categories": [
  "model_theft_extraction"
 ],
 "severity": 1,
 "targets": [
  "OpenAI",
  "open_weight_model"
 ],
 "products": [
  "GPT-2"
 ],
 "incident_date": "2019-08-22",
 "incident_date_precision": "day",
 "first_reported": "2020-10-23",
 "sources": [
  {
   "url": "https://atlas.mitre.org/studies/AML.CS0007",
   "publisher": "MITRE ATLAS",
   "title": "GPT-2 Model Replication",
   "published_at": "2020-10-23"
  },
  {
   "url": "https://wired.com/story/dangerous-ai-open-source",
   "publisher": "Wired",
   "title": "Wired Article, \"OpenAI Said Its Code Was Risky. Two Grads Re-Created It Anyway\"",
   "published_at": "2020-10-23"
  },
  {
   "url": "https://blog.usejournal.com/opengpt-2-we-replicated-gpt-2-because-you-can-too-45e34e6d36dc",
   "publisher": "blog.usejournal.com",
   "title": "Medium BlogPost, \"OpenGPT-2: We Replicated GPT-2 Because You Can Too\"",
   "published_at": "2020-10-23"
  }
 ],
 "source_count": 3,
 "confidence": 0.9,
 "status": "published",
 "origin": "seed_atlas",
 "tags": [
  "model-extraction",
  "replication",
  "gpt-2"
 ],
 "created_at": "2026-10-18T06:55:08.239Z",
 "updated_at": "2026-10-18T06:55:08.239Z"
}