Skip to content
NewA calmer, faster ZeroHour
StoryAI researchinfo

Anthropic Open-Sources FlashProve for Verifiable AI Outputs

3 sources · 3 articlesFirst reported Updated ()
What's new

2026-11-04: Anthropic Frontier Red Team with EPFL, MIT, StarkWare open-sources FlashProve prototype including verifiers, tested on Llama 4 Maverick MoE with performance metrics.

Merged summary · x-ai/grok-4.7 · rewritten as coverage arrivesSummaries by Vani AI

Anthropic open-sourced FlashProve, a zkSNARK system proving AI replies derive from committed weights, tested on Llama 4 Maverick.

Anthropic released FlashProve, using zkSNARK proofs to attest AI responses originate from committed model weights without exposure. The open-source prototype supports local verification, resists impersonation, and enables third-party claims like vulnerability scans. Collaborators include the Anthropic Frontier Red Team, EPFL, MIT, and StarkWare; it was tested on the Llama 4 Maverick 400B parameter MoE model.

Key facts
  • FlashProve generates zkSNARK proofs for AI output authenticity
  • Tested on Llama 4 Maverick 400B MoE requiring 80 GB GPU
  • Prover, C++ verifier, and Lean verifier released under Apache 2.0
  • Proof size ~18.78 MB for 1024-token requests
  • Verification takes ~17 seconds on 24 CPU threads
  • Offline commitment ~39 minutes per model
Explain this AI news with Vani AIWhat was announced, what is new and why it matters, in a new chat

Timeline

dated events in the reporting · oldest first
  1. Nov 3, 2026Anthropic releases FlashProve for cryptographic proofs of committed weights · MarkTechPost
  2. Nov 3, 2026Anthropic develops open-source system for proving AI model identity · TechCrunch
  3. Nov 4, 2026FlashProve prototype open-sources with test on Llama 4 Maverick MoE · Help Net Security

What each outlet adds

beyond the shared facts
  • MarkTechPostInitial announcement of FlashProve zkSNARK proofs for AI replies
  • TechCrunchDetails on GitHub release, local verification, and third-party claim support
  • Help Net SecurityPrototype technical specs, verification metrics, and test on Llama 4 Maverick
Vendors
Anthropic
Products
FlashProve
Organizations
EPFLMITStarkWareBroad Institute
AI models
AnthropicLlama 4 MaverickFlashProve

Coverage

oldest first · one row per article · UTC
  1. · 8h ago MarkTechPostFirst report

    Anthropic Releases FlashProve for Verifiable AI Replies

    Anthropic's FlashProve tool generates zkSNARK proofs to confirm AI responses derive from committed model weights.

  2. · 7h ago TechCrunch · AI

    Anthropic develops FlashProve for provable AI model identity

    An open-source system offering cryptographic proofs that outputs come from claimed models without revealing weights; lightweight and local verification possible.

  3. · 1h ago Help Net Security

    Anthropic open-sources FlashProve provable inference prototype

    FlashProve cryptographic proofs confirm specific models generated specific replies, tested on Llama 4 Maverick MoE model.