OpenMined is a 501(c)(3) nonprofit building open-source privacy-preserving AI infrastructure that enables secure computation across siloed data. Their tools allow AI auditors and researchers to evaluate proprietary AI systems without requiring direct access to sensitive models or data.
OpenMined is a 501(c)(3) nonprofit building open-source privacy-preserving AI infrastructure that enables secure computation across siloed data. Their tools allow AI auditors and researchers to evaluate proprietary AI systems without requiring direct access to sensitive models or data.
People
Updated 04/02/26Executive Director
Board Member
Research Team
Product Lead
Lead and Principal Engineer on BioVault.net
HR & People Operations Manager
Applied Researcher
Funding Details
Updated 04/02/26- Annual Budget
- $2,388,315
- Current Runway
- -
- Funding Goal
- -
- Funding Raised to Date
- $10,000,000
Org Details
Updated 04/02/26OpenMined was founded in 2017 by Andrew Trask, a deep learning researcher who recognized that the dominant paradigm of centralizing data for AI training was fundamentally incompatible with privacy, ownership, and accountability. The organization began as an open-source community and grew into a 501(c)(3) nonprofit foundation — formally incorporated in New York, NY — with a distributed team of over 50 employees and a contributor community of more than 16,000 researchers, engineers, and policy experts worldwide. The organization's core technical offering is PySyft, a Python library enabling secure and private machine learning through federated learning, differential privacy, multi-party computation, and homomorphic encryption. Building on this, OpenMined has developed SyftBox and SyftHub, platforms for governed federated AI computation that keep data distributed while allowing cross-institutional queries and audits. Their Attribution-Based Control framework enables data owners to maintain oversight of how their information is used even after it has been shared. A major focus area is enabling third-party AI audits of proprietary systems. Working with the UK AI Safety Institute and Anthropic, OpenMined demonstrated that secure enclaves (trusted execution environments) can be used to allow external researchers to evaluate frontier AI models without the model provider exposing weights, training data, or user logs. This work directly addresses a key gap in AI governance: the difficulty of independently verifying claims about AI capabilities and safety without access to proprietary systems. OpenMined has received funding from the Sloan Foundation, Georgetown University, Microsoft, Twitter/X, the Future of Life Institute ($1.66M in 2025), Open Philanthropy/Coefficient Giving, and the UK government. The organization co-launched and chairs the United Nations PET Lab, deploying on the UN Global Platform, and joined NIST's NAIRR Pilot as a launch partner. As of 2024, OpenMined reported annual revenue of approximately $6.1 million and total expenses of approximately $2.4 million, with net assets of $3.7 million.
Theory of Change
Updated 04/02/26OpenMined believes that a key precondition for AI safety and governance is the ability to audit and evaluate AI systems independently — without requiring AI developers to grant direct access to sensitive models, training data, or user information. By building open-source infrastructure for privacy-preserving computation (secure enclaves, federated learning, homomorphic encryption, multi-party computation), OpenMined enables governments, researchers, and civil society organizations to ask questions of AI systems and verify safety-relevant claims without the AI developers being able to hide or obscure dangerous capabilities. If this infrastructure is widely deployed and adopted, it creates the technical substrate for meaningful external oversight and accountability of AI systems — a structural safeguard against misuse, capability concealment, or misaligned behavior at scale.
Grants Received
Updated 04/02/26Projects– no linked projects
Updated 04/02/26Discussion
No comments yet. Be the first to share your thoughts.