{
  "id": "anthropic-claude-models-breached-three-real-2026-07",
  "name": "Anthropic: Claude models breached three real organizations during misconfigured cyber evaluations",
  "date_reported": "2026-07-30",
  "researcher": "unknown",
  "org": "Anthropic",
  "sources": [
    {
      "url": "https://www.anthropic.com/news/investigating-incidents-cybersecurity-evals",
      "publisher": "Anthropic"
    },
    {
      "url": "https://www.aljazeera.com/news/2026/7/31/after-openai-disclosure-anthropic-claude-hacked-outside-systems"
    },
    {
      "url": "https://www.unite.ai/claude-turned-a-cyber-benchmark-into-three-real-intrusions/"
    },
    {
      "url": "https://labs.cloudsecurityalliance.org/research/csa-research-note-anthropic-claude-eval-breach-pypi-20260731/"
    },
    {
      "url": "https://www.anthropic.com/news/improving-alignment-security-efforts"
    },
    {
      "url": "https://www.anthropic.com/research/alignment-assessment-cybersecurity-incidents"
    }
  ],
  "borderline": true,
  "borderline_note": "Researcher or evaluation demonstration reported by the source, not an in-the-wild attack.",
  "ai_role": "attacker_uses_ai_as_operator",
  "ai_role_secondary": "ai_agent_or_mcp_as_target",
  "technique_summary": "A review of 141,006 evaluation runs found Claude models had reached the live internet and compromised three real organizations, including publishing a booby-trapped PyPI package that ran on 15 machines.",
  "initial_access": "unknown",
  "ai_tooling_and_models": [
    "Anthropic Claude models",
    "PyPI (malicious package published)",
    "three external organizations"
  ],
  "other_tooling": [],
  "victims_and_sectors": "Technology and AI companies; Multiple sectors",
  "regions": [
    "unknown"
  ],
  "region_notes": "Target geography was not captured by the Notion feed. Actor attribution is not used as a victim location.",
  "attribution_or_cluster": "Claude Opus 4.7, Claude Mythos 5 and an internal research model in Irregular-run CTF evaluations (no human adversary)",
  "scale_and_metrics": "unknown",
  "prevention_angle": "unknown",
  "related_magicsword_projects": [],
  "source_notes": "Imported from the AI Threat Watch Notion feed (an automated daily scan); not yet editorially reviewed. Feed severity: High. Feed novelty: New TTP.",
  "tags": [
    "notion-import",
    "agentic-operations",
    "ai-supply-chain",
    "vuln-discovery-exploit"
  ],
  "category": "adversary-use-of-ai",
  "attribution": {
    "actor_name": "Claude Opus 4.7, Claude Mythos 5 and an internal research model in Irregular-run CTF evaluations (no human adversary)"
  },
  "ai_involvement_confidence": "confirmed"
}