named programs · 2 · from sources
What they call their work
PySyft Framework Development
Open-source framework enabling responsible, external access to sensitive data through privacy-preserving techniques like secure enclaves and structured transparency.
Secure Enclaves for AI Evaluation
Pilot experiment using NVIDIA H100 secure enclaves and OpenMined’s PySyft framework to conduct confidential AI evaluations across organizations, preserving privacy of both model and dataset.
activities · 3 groups
What they do
-
Privacy-Preserving Cybersecurity & Data Collaboration 9 activities
- Conducting applied research on secure computation and AI evaluationPerforms pilot studies and practical tests using secure enclaves (e.g., NVIDIA H100), OpenMined’s PySyft, and OpenDP to evaluate AI models against confidential datasets. Projects include collaboration with the UK AI Safety Institute and Anthropic, resulting in insights into current security limitations and future research directions for cross-organizational AI evaluation.
- Delivering education and training in privacy-enhancing technologiesOffers comprehensive training programs and educational courses—including the Private AI series (“Our Privacy Opportunity,” “Foundations of Private Computation,” “Introduction to Remote Data Science”)—to build expertise in privacy-preserving data science. Training is delivered to researchers in the NAIRR pilot and members of research consortia.
- Developing and maintaining open-source privacy-preserving frameworksDesigns, builds, and maintains open-source tools like PySyft and Syft to enable secure, governed computation on siloed data without requiring data to leave its original location. These frameworks support distributed data science, federated learning, and privacy-preserving data analysis across multiple domains.
- Developing privacy-preserving solutions for public interest research challengesDesigns and implements technical tools and pilot projects to demonstrate how privacy-enhancing technologies can enable responsible AI research and governance. Includes participation in the UK-US PETs Prize Challenge (winning Phase 1 with DeepMind) and developing analytics solutions for secure data collaboration in Phase 2.
- Enabling privacy-preserving cross-organizational AI research and auditsDeploys PySyft and other privacy-enhancing technologies to allow external researchers to study proprietary AI systems—such as recommendation algorithms at LinkedIn, Dailymotion, and Twitter—while protecting user privacy and intellectual property. This includes conducting remote audits and analyses without exposing sensitive data or models.
- Fostering global collaboration through open research communitiesProvides researchers with access to a global network of over 16,000 engineers and researchers to support knowledge exchange and collaborative development in privacy-preserving AI. Operates as a decentralized, community-driven effort to build a public network for non-public information.
- Operating federated research subnetworks for sensitive dataRuns BioVault, a free and open-source federated subnetwork that enables collaborative genomic research while keeping sensitive participant data on local devices. Also provides infrastructure and tools for launching similar privacy-preserving subnetworks in healthcare, climate, and finance.
- Providing technical infrastructure and support for research consortiaOffers software tools, secure servers, and ongoing technical support to enable research groups to conduct collaborative, privacy-preserving data science. Supports consortia in securely using each other’s private data without direct data sharing.
- Supporting international privacy-preserving data collaborationEnables secure, cross-border data analysis between national statistical agencies and international bodies using privacy-preserving techniques such as secure multiparty computation and differential privacy. Examples include data joins between Statistics Canada and the US Census Bureau, Statistics Canada and Istat, and a 2022 pilot between the U.S. Census Bureau and the UN PET Lab.
-
-
AI Governance and Ethical Research 1 activity
- Advancing responsible AI governance through policy engagementParticipates in high-level policy forums including the U.S. Senate AI Insight Forum and events hosted by the White House OSTP, NSF, and PCAST, contributing expertise on AI transparency, explainability, intellectual property, and privacy-preserving methods.
-
-
Human Rights Advocacy Funding 1 activity
- Funding research on algorithmic accountability with privacy-enhancing technologiesProvides financial support for research projects that use privacy-preserving methods to study the societal impact of AI-powered recommendation engines, aiming to improve algorithmic transparency and accountability.
-
financials · form 990 · fy2024
revenue
Total revenue$6.10M
Contributions & grants$6.09M100%
Program service revenue$00%
Investment income$7K0%
Other revenue$0
expenses
Total expenses$2.39M
Program expenses81%
Admin / overhead10%
Fundraising9%
Salaries & benefits$103K
Grants paid out$0
Largest expense lineProfessional Fees
balance sheet
Total assets$3.72M
Cash$3.55M
Investments$0
Liabilities$9K
Net assets$3.71M
Liquid reserves17.8 mo
1 years on record · 2024–2024
leadership · form 990 part vii · fy2024
Who runs it
paid leadership · 5
| Name | Title | Hours/wk | Compensation |
|---|---|---|---|
| RONNIE FALCON | TREASURER, CHIEF PRODUCT OFFICER | 40 | $192K |
| MADHAVA JAY | HEAD OF ENGINEERING | 40 | $143K |
| PETER SMITH | CHIEF FINANCIAL OFFICER | 40 | $133K |
| BENNETT FARKAS | CHIEF MARKETING OFFICER | 40 | $96K |
| LACEY STRAHM | HEAD OF POLICY | 40 | $58K |
board members · 2
- ANDREW TRASK — PRESIDENT/EXECUTIVE DIRECTOR
- ROD DEWAR — SECRETARY
relationships · 33
Who they work with
- Anthropic Partner — Partnered with OpenMined to test secure enclaves for AI evaluation, providing a proxy model for the experiment.
- BioVault Partner — OpenMined operates BioVault as a dedicated subnetwork for secure, collaborative genomic research.
- Christchurch Call Initiative on Algorithmic Outcomes Partner — Collaborated on a technical report demonstrating privacy-preserving auditing of social media algorithms using OpenMined technologies.
- Christchurch Call Initiative on Algorithmic Outcomes Partner — Collaborated with OpenMined to deploy PySyft for external oversight of AI systems on platforms like Dailymotion and LinkedIn.
- Dailymotion Partner — Collaborated on privacy-preserving audits of its recommendation algorithms using OpenMined's PySyft and OpenDP.
- Dailymotion Partner — Platform where OpenMined deployed PySyft to enable privacy-preserving external oversight of AI systems.
- DeepMind Partner — Collaborating with OpenMined on a privacy-preserving data analytics solution for the UK-US PETs Prize Challenge.
- Department for Science, Innovation, and Technology Government — Profiled OpenMined's PySyft in its portfolio of AI assurance techniques.
- Government of New Zealand Government — Stakeholder in the Christchurch Call Initiative collaborating on algorithmic accountability research.
- Government of the United States Government — Stakeholder in the Christchurch Call Initiative collaborating on algorithmic accountability research.
- Italian National Institute of Statistics Partner — Collaborated on a privacy-preserving data sharing pilot using PySyft to perform cross-border data joins.
- LinkedIn Partner — Collaborated on privacy-preserving audits of its recommendation algorithms using OpenMined's PySyft and OpenDP.
- LinkedIn Partner — Platform where OpenMined deployed PySyft to enable privacy-preserving external oversight of AI systems.
- Microsoft Partner — Collaborates on research into privacy-preserving AI for the Christchurch Call Initiative.
- NAIRR Partner — Launch partner in the National Artificial Intelligence Research Resource (NAIRR) pilot program.
- National Institute of Standards and Technology Government — Partnering with OpenMined through participation in the UK-US Privacy-Enhancing Technologies (PETs) Prize Challenge.
- National Science Foundation Government — Non-governmental partner in the NAIRR pilot program launched by the NSF.
- President’s Council of Advisors on Science and Technology Government — Engaged in event co-hosted by PCAST on advancing AI research for public benefit.
- Statistics Canada Partner — Collaborated on a privacy-preserving data sharing pilot using PySyft to perform cross-border data joins.
- Twitter Partner — Collaborates with OpenMined to test privacy enhancing technologies for third-party access to non-public data.
- Twitter Partner — Collaborates with Twitter’s ML Ethics, Transparency, and Accountability (META) team to enable privacy-preserving research into Twitter’s algorithms and datasets.
- Twitter Partner — Participates in a multistakeholder initiative with OpenMined to study online content recommendation systems.
- U.S. Census Bureau Partner — Collaborated with OpenMined through its emerging technologies group xD in a pilot using PySyft for international data collaboration.
- U.S. National Science Foundation Government — Participated in joint event with NSF on AI research opportunities and resource access.
- U.S. National Science Foundation Partner — Collaborates with NSF on the NAIRR pilot program to democratize access to AI research resources.
- U.S. Senate AI Working Group Partner — Collaborated through participation in the Senate's AI Insight Forums on AI policy development.
- UK AI Safety Institute Partner — Collaborated with OpenMined on a pilot experiment to test secure enclaves for AI evaluation using confidential datasets.
- UK Frontier AI Taskforce Partner — Collaborates to develop technical infrastructure for AI safety research.
- UN PET Lab Partner — Partnered with OpenMined to pilot privacy-preserving data collaboration using PySyft across member countries and the United Nations.
- United Nations Statistics Division Partner — Partnered through the UN PET Lab to support international privacy-preserving data sharing collaboration.
+ 3 more
strategies · 3
How they approach the work
Named approaches extracted from this org’s sources. Where others share an approach, follow it to see the full set of orgs running it.
- Cross-Sector Partnership Model for Responsible AImethodology: cross_sector_collaborationBy fostering collaboration across government, academia, industry, and nonprofits, we amplify the impact of AI research on societal challenges, because diverse partnerships combined with privacy-preserving infrastructure lower barriers to entry and align innovation with public interest.
- Privacy-Preserving Collaborative Infrastructuremethodology: privacy_preserving_collaborationBy enabling secure computation and collaboration on non-public data without centralization, we produce broader access to sensitive datasets for AI research and societal benefit, because decentralized governance and privacy-preserving technologies (e.g., federated computation, secure enclaves, structured transparency) allow data owners to retain control while permitting governed analysis.
- Structured Transparency for Algorithmic Governancemethodology: structured_transparencyBy applying structured transparency—enforcing input, output, and runtime policies—we enable external audits and replication of AI systems without exposing sensitive data or models, because this governance framework balances accountability with privacy, security, and intellectual property protection.