{
  "id": 17669,
  "url": "https://www.forrester.com/blogs/four-ai-escapes-just-redefined-responsible-ai",
  "title": "Four AI Escapes Just Redefined “Responsible AI”",
  "summary": "In nine days this month, OpenAI and Anthropic disclosed that their own models escaped safety evaluations and reached production systems at four other companies. No model malfunctioned — the governance did. Responsible AI now has to govern what agents do, not only what models decide, and that takes a responsible AI deployment policy.",
  "authors": "Brian Hopkins",
  "category": "org",
  "topics": "regulation,agents-autonomy",
  "orgs": "openai,anthropic",
  "regions": null,
  "published_at": "2026-08-07T19:35:52.000Z",
  "fetched_at": "2026-08-08T05:10:34.355Z",
  "source_slug": "x-forrester-ai-blog",
  "source_name": "Forrester AI blog",
  "source_homepage": "https://www.forrester.com/blogs/category/artificial-intelligence-ai/",
  "ethics_ai_record_url": "https://ethics.ai/record/17669",
  "original_url": "https://www.forrester.com/blogs/four-ai-escapes-just-redefined-responsible-ai",
  "evidence_status": "source-only",
  "attribution": "via ethics.ai"
}