NEW YORK – June 3, 2026 – Lagrange Labs today announced the open-source release of DeepProve, its production-grade zero-knowledge machine learning (zkML) system. DeepProve has generated more than 12 million cryptographic proofs and verified more than 3 million AI inferences end-to-end over the past year. The full stack — circuits, prover, verifier, and ONNX pipeline — is now freely available to developers, researchers, and enterprises worldwide.
Background and Market Context
AI systems have moved beyond demos into the critical infrastructure underlying finance, defense, healthcare, and enterprise operations. As AI agents take real-world actions and models issue consequential recommendations, the ability to verify what actually happened has become essential.
The urgency is reflected in recent data: Stanford’s 2026 AI Index documented 362 AI incidents in 2025, up 55 percent year-over-year. GitHub’s 2026 State of AI Code found that 46 percent of new code is now AI-generated, while verification practices have not kept pace. McKinsey research indicates 71 percent of enterprise executives say they will not scale AI systems without proof of correctness. The EU AI Act’s high-risk AI obligations take effect on August 2, 2026.
What DeepProve Does
DeepProve gives every AI inference a cryptographic receipt: a verifiable proof that the correct model ran on the correct input and produced the output being presented. Proofs can be verified in milliseconds, on-chain if desired, without exposing private model weights or input data.
What Is Included in the Open-Source Release
The repository includes:
— The complete proving system, including circuits, prover, and verifier. — Native support for ONNX, safetensors, and GGUF formats, enabling models from PyTorch, TensorFlow, JAX, or Hugging Face to run through DeepProve unchanged. No other zkML library supports the formats large language models actually ship in. — Reproducible, layer-by-layer benchmarks. DeepProve generates LLM proofs up to 60× faster than the prior state-of-the-art and achieves 671× faster verification — while fully preserving model accuracy. — Working examples across GPT-2, Gemma-3, classifiers, and vision models, with Llama-class models in active development. — A public dashboard at deep-prove.lagrange.dev showing live proofs running in production. — A research paper documenting the methodology behind the performance and accuracy results. The repository is available at: https://github.com/Lagrange-Labs/deep-prove
Production Track Record
DeepProve has operated in production for twelve months since the first full-LLM proof was generated in August 2025. Lagrange Labs is working with partners including Anduril, IBM, Qualcomm, Lockheed Martin, Oracle, Intel, NVIDIA, and AWS, as well as more than 200 crypto and enterprise ecosystem partners across the Lagrange network. The company has maintained an ongoing compliance engagement with the U.S. Securities and Exchange Commission regarding the role of cryptographic verification in AI systems.
DeepProve served as the proof layer behind Turing Roulette, a live public demonstration in which more than 500,000 unique participants generated 3.7 million live AI inferences, each cryptographically verified.
Quote from Leadership
“We didn’t build DeepProve to own verifiable AI. We built it so nobody had to. AI needs a verification layer, and that layer should be open. Today, DeepProve becomes a primitive. The black box is open.” – Ismael Hishon-Rezaizadeh, Founder and CEO, Lagrange Labs
Availability
The DeepProve repository, live dashboard, and research paper are available now at https://github.com/Lagrange-Labs/deep-prove and deep-prove.lagrange.dev. Developers building agentic infrastructure can attach a cryptographic receipt to every model call. Teams deploying AI in regulated environments can attach verifiable evidence to each AI decision. Researchers have access to a production-grade zkML system to inspect, benchmark, and extend.
About Lagrange Labs
Lagrange Labs builds verifiable infrastructure for AI and crypto. Its DeepProve system provides cryptographic proofs for AI inferences, enabling anyone to verify that an AI model produced a specific output for a specific input without revealing the model or the input. Lagrange Labs is headquartered in New York City.
Media ContactCompany Name: Lagrange LabsContact Person: Isabel ChiEmail: Send EmailCountry: United StatesWebsite: https://lagrange.dev/
