{
 "id": "AIR-2020-0029",
 "slug": "model-extraction-and-adversarial-attacks-on-translation-apis",
 "title": "Model Extraction and Adversarial Attacks on Translation APIs",
 "summary": "Researchers replicated commercial machine translation services via public APIs and transferred adversarial inputs to trigger targeted errors. The demonstration proved that black-box models are vulnerable to extraction and evasion.",
 "type": "attack_on_ai",
 "categories": [
  "model_theft_extraction",
  "adversarial_evasion"
 ],
 "severity": 1,
 "targets": [
  "Google",
  "Microsoft"
 ],
 "products": [
  "Google Translate",
  "Bing Translator",
  "Systran Translate"
 ],
 "incident_date": "2020-04-30",
 "incident_date_precision": "day",
 "first_reported": "2020-11-18",
 "sources": [
  {
   "url": "https://atlas.mitre.org/studies/AML.CS0005",
   "publisher": "MITRE ATLAS",
   "title": "Attack on Machine Translation Services",
   "published_at": "2020-11-18"
  },
  {
   "url": "https://arxiv.org/abs/2004.15015",
   "publisher": "arXiv",
   "title": "Wallace, Eric, et al. \"Imitation Attacks and Defenses for Black-box Machine Translation Systems\" EMNLP 2020",
   "published_at": "2020-11-18"
  },
  {
   "url": "https://ericswallace.com/imitation",
   "publisher": "ericswallace.com",
   "title": "Project Page, \"Imitation Attacks and Defenses for Black-box Machine Translation Systems\"",
   "published_at": "2020-11-18"
  },
  {
   "url": "https://thehill.com/policy/international/asia-pacific/449164-google-under-fire-for-mistranslating-chinese-amid-hong-kong",
   "publisher": "thehill.com",
   "title": "Google under fire for mistranslating Chinese amid Hong Kong protests",
   "published_at": "2020-11-18"
  }
 ],
 "source_count": 4,
 "confidence": 0.95,
 "status": "published",
 "origin": "seed_atlas",
 "tags": [
  "model-extraction",
  "adversarial-evasion",
  "translation-api"
 ],
 "created_at": "2026-10-18T06:55:08.239Z",
 "updated_at": "2026-10-18T06:55:08.239Z"
}