EZKL
An engine for converting ML models into verifiable zero-knowledge proofs
Performances
RD Popularity Index
RD Growth Index
Comparison
Team
Follow Updates
Follow Lists
People
About EZKL
EZKL is a developer-focused toolkit that turns deep learning models into zero-knowledge proofs. Its core business is to let users prove that an AI model ran correctly on private data without revealing the data or the model itself. The main value proposition is bridging AI and cryptography: it converts standard ONNX graphs into Halo2-based SNARK circuits, enabling verifiable inference on-chain, in browsers, or on edge devices.
The market pain point it addresses is the trust gap in AI: today, users cannot verify that a model's output is genuine, and data owners are reluctant to share sensitive inputs. EZKL solves this by providing cryptographic guarantees of computation integrity while preserving privacy. Its competitive edge lies in its ease of use—no circuit design expertise is required—and its practical deployment options, including Solidity verifiers and JavaScript tooling.
In the past six months (2026-02-28 to 2026-08-31), EZKL has focused on scaling proof generation through Lilith, a cloud-based service that offloads heavy computation. It also introduced collaborative SNARKs (co-SNARKs), allowing two parties to jointly generate a proof without revealing their private inputs to each other—a significant step for multi-party AI scenarios. The project has maintained its audited codebase and continued to expand its ecosystem integrations. One notable limitation remains: proof generation is still computationally expensive for very large models, and the tooling requires a certain level of familiarity with cryptographic concepts, though the team is actively working to lower this barrier.
Major progress in the past six months (2026-03-01 to 2026-09-01):
Launched Collaborative SNARKs (cosnarks), enabling two or more parties to jointly generate zero-knowledge proofs without revealing private data to each other. This is led by Halo2 co-author Ying Tong Lai and is now integrated into EZKL, allowing reuse of existing proving/verification keys with minimal workflow changes.
Integrated cosnark proving into Lilith, the cloud-scale proof generation cluster, enabling quick setup of collaborative proof peer servers via a single command.
Released cosnarks in closed beta, with trials underway for early adopters.
Future development focus and roadmap:
Introducing asymmetric MPC to reduce compute burden on data providers (e.g., mobile devices), complementing iOS bindings and metal acceleration. This is an active research area and will be explored in the coming months.
Expanding beyond Halo2 to incorporate additional proving systems for broader flexibility.
Continuing to optimize proof generation for large models and high-throughput scenarios, with GPU acceleration and folding schemes expected to reduce proof sizes to kilobytes by 2027.
Jason Morton (CEO)
Jason Morton is the co-founder and CEO of EZKL. He previously served as a tenured Mathematics and Statistics professor, received the DARPA Young Faculty Award for quantum-inspired cryptographic algorithms, and was a PI in the original DARPA Deep Learning program in 2011. He founded and sold an Ethereum-based financial intermediary, served as Head of Cryptography for Lit Protocol, and holds an A.B. in Psychology from Harvard, an M.A. in Economics from the University of Michigan, and a Ph.D. in Mathematics from U.C. Berkeley.