{
  "id": 8192724,
  "title": "How OpenAI Decided Astra Was Too Dangerous to Ship Open: Capability Thresholds, Explained",
  "url": "https://urgent.news/2026/09/18/how-openai-decided-astra-was-too-dangerous-to-ship-open-capability",
  "topic": "ai",
  "section": "AI",
  "published": "2026-09-18T06:43:37.000Z",
  "source": {
    "name": "Dev.to",
    "slug": "dev-to",
    "url": "https://dev.to/muskan_bandta/how-openai-decided-astra-was-too-dangerous-to-ship-open-capability-thresholds-explained-bj1"
  },
  "original_language": "en",
  "account": "OpenAI's decision to rate the Astra model \"Critical\" for cybersecurity and gate its dangerous capabilities was a result of a structured process for determining when a model is too dangerous to release without safeguards. This process focuses on capability thresholds in specific domains, such as cybersecurity, rather than just asking if the model is safe. Astra's ability to autonomously find zero-day vulnerabilities in hardened systems and turn them into working exploits triggered this threshold, which in turn prompted OpenAI to implement specific safeguards before shipping the model. The framework of defining threshold ahead of time, evaluating honestly, and gating dangerous capabilities is meant to prevent ad-hoc judgments during launch pressure. Understanding this process is important because it becomes a signal about new existing capabilities in the world, and organizations deploying AI systems will eventually need to define their own versions of this framework for governance.",
  "summary": "When GPT-6 Astra shipped on September 3, the headline was that OpenAI rated it Critical for cybersecurity, the first time it has ever given a model that rating, and chose to gate the dangerous capability rather than release it openly. A lot of coverage treated \"Critical\" as a vibe. It is not. It is the output of a structured process for deciding when a model is too dangerous to ship as-is, and…",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}