{
  "id": 18153733,
  "title": "Progress on provable inference for large language models",
  "url": "https://urgent.news/2026/11/04/progress-on-provable-inference-for-large-language-models",
  "topic": "ai",
  "section": "AI",
  "published": "2026-11-04T17:17:46.000Z",
  "source": {
    "name": "Hacker News",
    "slug": "hacker-news",
    "url": "https://www.anthropic.com/research/provable-inference"
  },
  "original_language": "en",
  "account": "Anthropic, EPFL, MIT and multiple researchers have made progress on a type of cryptographic proof called \"zkSNARK\" for large language models. Currently, when a user talks to an AI model, they have no way of knowing for certain that the responses came from the specific model claimed by developers. This is because most AI model weights are kept private. Provable inference aims to enable a user with limited resources to verify that a response came from the requested model without running the model itself. The researchers released FlashProve, an open-source prototype for provable inference that can generate proofs at over 100 tokens per second for a 400B-parameter model using one H100 GPU. The release includes code, Lean formalization, and a technical description of the protocol.",
  "summary": null,
  "key_points": [
    "Researchers from Anthropic, EPFL, MIT advance zkSNARK proofs for LLMs",
    "FlashProve prototype generates proofs at >100 tokens/second for 400B model",
    "Open-source release includes code, Lean formalization, technical description"
  ],
  "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."
}