{
  "id": 13457,
  "url": "https://www.darkreading.com/cybersecurity-operations/incorrigible-ai-models-resist-rehabilitation",
  "title": "Escape Artists: 'Incorrigible' AI Models Resist Rehabilitation",
  "summary": "The hacking of Hugging Face by a rogue OpenAI agent is significant, but unsurprising — and preventing the next AI model escape will be difficult, at best.",
  "authors": "Robert Lemos",
  "category": "news",
  "topics": "agents-autonomy",
  "orgs": "openai,huggingface",
  "regions": null,
  "published_at": "2026-07-24T19:45:02.000Z",
  "fetched_at": "2026-07-25T05:10:48.796Z",
  "source_slug": "x-dark-reading-ai-security",
  "source_name": "Dark Reading (AI security)",
  "source_homepage": "https://www.darkreading.com",
  "ethics_ai_record_url": "https://ethics.ai/record/13457",
  "original_url": "https://www.darkreading.com/cybersecurity-operations/incorrigible-ai-models-resist-rehabilitation",
  "evidence_status": "source-only",
  "attribution": "via ethics.ai"
}