{
 "license": "CC-BY-4.0",
 "attribution": "Fide AI, Agentic Cyber Explorer",
 "url": "https://agentic-cyber-explorer.pages.dev/findings/aixcc-systems-patched-most-found-bugs/",
 "asOf": "2026-09-26",
 "id": "aixcc-systems-patched-most-found-bugs",
 "claim": "In DARPA's AI Cyber Challenge, autonomous systems patched most of the synthetic vulnerabilities they found.",
 "evidenceKind": "measured",
 "scope": "Competition challenges; DARPA's own scoring.",
 "topics": [
  "vulnerability-repair"
 ],
 "atlas": [],
 "evidence": [
  {
   "event": "darpa-aixcc-semifinal-results-2024"
  },
  {
   "event": "darpa-aixcc-final-results-2025",
   "note": "43 of 54 found synthetic vulnerabilities patched."
  }
 ],
 "relations": [],
 "statusHistory": [
  {
   "status": "reported",
   "on": "2024-08-11",
   "why": "Semifinal results.",
   "event": "darpa-aixcc-semifinal-results-2024",
   "kind": "evidence"
  },
  {
   "status": "qualified",
   "on": "2026-02-07",
   "why": "A later review finds 16-21% of top patches semantically wrong.",
   "event": "aixcc-sok-competition-lessons-2026",
   "kind": "evidence"
  },
  {
   "status": "qualified",
   "on": "2026-09-25",
   "why": "Correction: the 16-21% figure is competition-scored submission accuracy and does not reduce DARPA's 43 counted patches. The qualification now rests on PatchBench: agents from top AIxCC teams lose much of their solve rate under stronger-than-crash validation.",
   "event": "patchbench-vulnerability-patching-validity-2026",
   "kind": "correction"
  }
 ],
 "halfLifeDays": 540,
 "wouldChange": "Independent verification of competition patches.",
 "fideQuestions": [
  "FID-088"
 ],
 "methods": [
  "ai-vulnerability-discovery",
  "automated-patching"
 ],
 "review": "assistant-drafted",
 "addedOn": "2026-09-25"
}