{
  "type": "post",
  "stable_id": "post:shieldstral-inference-time-safety-policy",
  "slug": "shieldstral-inference-time-safety-policy",
  "title": "Shieldstral Turns Guardrails Into An Inference-Time Policy",
  "description": "Mistral released a 3B open-weights multimodal safety classifier whose policy is supplied as a natural-language question at inference time.",
  "retrieval_nugget": "Shieldstral converts moderation into instruction-query-document binary QA and returns a calibrated yes/no score. It is Apache 2.0, handles text and images, and is designed to run on one 16 GB GPU; benchmark claims are vendor-reported.",
  "published_at": "2026-08-04",
  "updated_at": "2026-08-05",
  "record_date": "2026-08-04",
  "date_kind": "published_at",
  "topics": [
    "ai-safety",
    "multimodal",
    "open-weights",
    "guardrails",
    "model-routing"
  ],
  "entities": [
    "Mistral AI",
    "Shieldstral"
  ],
  "source_urls": [
    "https://mistral.ai/news/shieldstral/",
    "https://huggingface.co/mistralai/Shieldstral-1.0-3B",
    "https://arxiv.org/abs/2607.25857"
  ],
  "source_format": "article",
  "editorial_timing": {
    "lane": "urgent_half_hour",
    "scheduled_at": "2026-08-05T18:30:00+03:00",
    "real_news_delta": "primary-source release or incident report"
  },
  "visual_decision": {
    "status": "included",
    "reason": "the inference-time policy interface and its boundary from application enforcement are clearer as a diagram",
    "reviewed_by": "codex"
  },
  "schema_version": "newruntime-agent-readable-v0.2",
  "status": "published",
  "visuals": [
    {
      "role": "hero",
      "src": "/images/drip/shieldstral-inference-time-safety-policy/shieldstral-inference-time-safety-policy.webp",
      "alt": "A whiteboard policy diagram showing text, images, and an inference-time policy flowing through a small guard model into a safety score and application gate.",
      "caption": "New Runtime synthesis from Mistral's Shieldstral release."
    }
  ],
  "routes": {
    "html": "https://newruntime.com/posts/shieldstral-inference-time-safety-policy/",
    "markdown": "https://newruntime.com/posts/shieldstral-inference-time-safety-policy.md",
    "json": "https://newruntime.com/posts/shieldstral-inference-time-safety-policy.json"
  },
  "telegram_message_id": 2908,
  "telegram_url": "https://t.me/qwgai/2908",
  "telegram_message_ids": [
    2908,
    2909
  ],
  "telegram_delivery_mode": "text_then_media",
  "telegram_media_url": "https://t.me/qwgai/2909"
}