{"version":1,"type":"story","url":"https://digestai.news/story/anthropic-releases-flashprove-for-cryptographic-ai-inference-proofs","json":"https://digestai.news/story/anthropic-releases-flashprove-for-cryptographic-ai-inference-proofs.json","markdown":"https://digestai.news/story/anthropic-releases-flashprove-for-cryptographic-ai-inference-proofs.md","slug":"anthropic-releases-flashprove-for-cryptographic-ai-inference-proofs","headline":"Anthropic releases FlashProve for cryptographic AI inference proofs","summary":"Anthropic released FlashProve on November 2, 2026, an open-source research prototype that uses zero-knowledge proofs to verify a language model response came from committed weights without revealing them. The system targets a verification gap in hosted AI where users cannot confirm which model actually generated a response.\n\nFlashProve uses zkSNARKs in a software-only design. During an offline phase, the provider publishes a short commitment identifying exact weights while keeping them secret. For completed prompt-response pairs, the system produces a proof that output tokens match the committed model's output. Anyone with the commitment, prompt, and reply can verify without accessing weights. The implementation leverages Meta's Llama 4 Maverick's mixture-of-experts structure, with prover work scaling to roughly 17 billion active parameters per token rather than all 400 billion.\n\nAnthropic evaluated on a 4-bit conversion of Maverick using one H100 and 24 CPU cores. A batch of 24 requests took about 223 seconds to prove, producing 18.78 MB proofs each. The offline commitment step took about 39 minutes. Verification runs on CPU without GPU: 17 seconds with 24 threads, 21 seconds with one thread. Proving required about 72 times more per-GPU time than FP8 inference on eight H100s running vLLM. The repository includes a prover, C++ and Lean verifiers, and formal proofs under Apache 2.0, labeled version 0.16.0 as an unmaintained research artifact. Known limits include 1,024-token contexts, no independent security audit, and coverage only for exact integer weights under greedy decoding.","keyPoints":["FlashProve uses zkSNARKs to prove model outputs match committed weights without revealing them","Proving takes ~72x more GPU time than standard inference on H100s","Repository labeled v0.16.0 as unmaintained research artifact under Apache 2.0"],"whyItMatters":"FlashProve addresses a core trust problem in hosted AI: verifying which model actually generated a response. The high proving cost and current limits show the gap between cryptographic verification and practical deployment.","category":{"slug":"research","name":"Research","url":"https://digestai.news/category/research"},"entities":{"companies":["Anthropic","Meta"],"models":["Llama 4 Maverick","FlashProve"],"people":[]},"firstPublishedAt":"2026-11-03T04:44:05Z","updatedAt":"2026-11-03T04:44:05Z","sourceCount":1,"hasPrimarySource":false,"sources":[{"outlet":"Unite.AI","title":"Anthropic Releases FlashProve for Cryptographic AI Inference Proofs","url":"https://unite.ai/anthropic-releases-flashprove-for-cryptographic-ai-inference-proofs","publishedAt":"2026-11-03T04:44:05Z","type":"press","primary":false,"lead":true}],"sourceNotes":null,"discussions":[],"thread":null,"cite":{"text":"Digest AI, \"Anthropic releases FlashProve for cryptographic AI inference proofs\", 3 November 2026, https://digestai.news/story/anthropic-releases-flashprove-for-cryptographic-ai-inference-proofs","publisher":"Digest AI","title":"Anthropic releases FlashProve for cryptographic AI inference proofs","datePublished":"2026-11-03T04:44:05Z","url":"https://digestai.news/story/anthropic-releases-flashprove-for-cryptographic-ai-inference-proofs"},"generatedBy":"Written by Digest AI's editorial model from the linked sources; the sources are the record.","license":"Headlines, digests and key points are written by Digest AI and may be quoted with a link to the story page. Linked articles belong to their publishers. Terms: https://digestai.news/terms#reuse"}