{
 "id": "AIR-2026-0352",
 "slug": "rce-vulnerability-in-microsoft-semantic-kernel-search-plugin",
 "title": "RCE Vulnerability in Microsoft Semantic Kernel Search Plugin",
 "summary": "Researchers discovered a path in Microsoft Semantic Kernel where a single prompt injection could lead to host-level remote code execution. The vulnerability affects agents using the Search Plugin backed by the In-Memory Vector Store.",
 "type": "attack_on_ai",
 "categories": [
  "prompt_injection",
  "infra_vuln",
  "agent_hijack"
 ],
 "severity": 1,
 "targets": [
  "Microsoft",
  "framework_or_tooling"
 ],
 "products": [
  "Semantic Kernel"
 ],
 "incident_date": "2026-05-07",
 "incident_date_precision": "day",
 "first_reported": "2026-02-19",
 "sources": [
  {
   "url": "https://github.com/advisories/GHSA-xjw9-4gw8-4rqx",
   "publisher": "GitHub Advisory Database",
   "title": "CVE-2026-26030: Microsoft Semantic Kernel InMemoryVectorStore filter functionality vulnerable to remote code execution",
   "published_at": "2026-02-19"
  },
  {
   "url": "https://atlas.mitre.org/studies/AML.CS0062",
   "publisher": "MITRE ATLAS",
   "title": "RCE Vulnerability in Semantic Kernel Search Plugin",
   "published_at": "2026-06-30"
  },
  {
   "url": "https://microsoft.com/en-us/security/blog/2026/05/07/prompts-become-shells-rce-vulnerabilities-ai-agent-frameworks",
   "publisher": "Microsoft Security",
   "title": "When prompts become shells: RCE vulnerabilities in AI agent frameworks",
   "published_at": "2026-06-30"
  },
  {
   "url": "https://cve.org/CVERecord?id=CVE-2026-26030",
   "publisher": "cve.org",
   "title": "CVE-2026-26030",
   "published_at": "2026-06-30"
  }
 ],
 "source_count": 4,
 "confidence": 0.95,
 "status": "published",
 "origin": "seed_atlas",
 "tags": [
  "semantic-kernel",
  "rce",
  "microsoft",
  "CVE-2026-26030"
 ],
 "created_at": "2026-10-18T06:55:08.239Z",
 "updated_at": "2026-10-18T07:00:14.101Z"
}