{
 "id": "AIR-2023-0017",
 "slug": "azure-red-team-demonstrates-adversarial-evasion-on-microsoft",
 "title": "Azure Red Team demonstrates adversarial evasion on Microsoft edge AI product",
 "summary": "The Azure Red Team conducted an automated red team exercise that continuously manipulated target images to induce misclassifications in a Microsoft edge AI product. The finding highlights adversarial evasion risks in machine learning models.",
 "type": "attack_on_ai",
 "categories": [
  "adversarial_evasion"
 ],
 "severity": 1,
 "targets": [
  "Microsoft"
 ],
 "products": [
  "Microsoft Edge AI Product"
 ],
 "incident_date": "2023-03-31",
 "incident_date_precision": "day",
 "first_reported": "2023-03-31",
 "sources": [
  {
   "url": "https://github.com/avidml/avid-db/blob/main/vulnerabilities/2023/AVID-2023-V011.json",
   "publisher": "AVID",
   "title": "AVID-2023-V011: The Azure Red Team performed a red team exercise on a new Microsoft product designed for running AI workloads at the edge.",
   "published_at": "2023-03-31"
  },
  {
   "url": "https://atlas.mitre.org/studies/AML.CS0011",
   "publisher": "MITRE ATLAS",
   "title": "Microsoft Edge AI Evasion",
   "published_at": "2023-03-31"
  }
 ],
 "source_count": 2,
 "confidence": 0.85,
 "status": "published",
 "origin": "seed_avid",
 "tags": [
  "adversarial-evasion",
  "red-team",
  "edge-ai"
 ],
 "created_at": "2026-10-18T06:56:23.100Z",
 "updated_at": "2026-10-18T06:56:23.100Z"
}