Back to Gap Map Model components / Federated learning

Syft / PySyft

OpenMined
open source / Overall score: 2.5

PySyft is OpenMined's library for privacy-preserving data science on data that stays with its owner. A data scientist submits computations that the data owner approves and runs on private data, over cloud storage the organizations already use, such as Google Drive, with sandboxed job execution, per-collaborator approval and an optional Trusted Execution Environment package.

Openness

5 high confidence
5.0
license
Apache-2.0(OSI permissive)
source
public
core features withheld
no — every runtime piece is published - eleven syft-* packages under packages/ plus syft_client under one Apache-2.0 license - and PySyft runs on cloud storage the user already has rather than on an OpenMined-operated plane. No enterprise path and no license key in the tree

PySyft is Apache-2.0 licensed across the whole codebase, and nothing is withheld behind a paywall or license key: OpenMined publishes each of its runtime components under that same license. The product is also architected so there is no central OpenMined-run service to withhold in the first place, since computations run through cloud storage the user already controls.

  • https://api.github.com/repos/OpenMined/PySyft/contents/packages recorded 2026-08-12

    The published component set: enclave-model-api-example, syft-bg, syft-datasets, syft-enclave, syft-job, syft-migration, syft-notebook-ui, syft-permissions, syft-perms, syft-rds, syft-restrict. Together with syft_client at the root this is the whole runtime, under one Apache-2.0 LICENSE, with no ee/, enterprise/ or second-licensed directory anywhere in the tree.

  • https://github.com/OpenMined/PySyft recorded 2026-06-22

    Apache-2.0 license, public source, 9.9k stars

  • https://pypi.org/project/syft/ recorded 2026-06-22

    PyPI metadata lists License Apache-2.0, source OpenMined/PySyft

  • https://pypi.org/pypi/syft-crypto-python/json recorded 2026-09-28

    PyPI metadata for syft-crypto-python, the one OpenMined runtime dependency built outside the PySyft tree, reads "Python bindings for the Syft crypto protocol" with the classifier "License :: OSI Approved :: Apache Software License".

  • https://raw.githubusercontent.com/OpenMined/PySyft/dev/LICENSE recorded 2026-09-28

    The root LICENSE file on the dev branch is the standard "Apache License" text, "Version 2.0, January 2004".

  • https://raw.githubusercontent.com/OpenMined/PySyft/dev/pyproject.toml recorded 2026-09-28

    pyproject.toml describes the package as "privacy-preserving data science over file-based transports" and lists runtime dependencies including Google API clients, workspace syft packages and "syft-crypto-python>=0.1.2b2"; searched for an enterprise extra or license-key dependency: none.

  • https://raw.githubusercontent.com/OpenMined/PySyft/dev/README.md recorded 2026-09-28

    README carries the badge "License: Apache 2.0" and the closing link "Apache License 2.0", and says computations run "all through cloud storage their organizations already use" with "No new infrastructure required." Its closing section says "OpenMined is a non-profit foundation". Searched the body for enterprise, pricing, paid and license key: none.

  • https://ungh.cc/repos/OpenMined/PySyft/files/dev recorded 2026-09-28

    The public file listing of the dev branch carries the workspace packages, among them "packages/syft-enclave/README.md" and "packages/syft-permissions/pyproject.toml", beside one root "LICENSE". Searched every path for ee/, enterprise/, commercial/ and a second LICENSE file: none.

Adoption

1 medium confidence
1.0

Adoption is measured by PyPI downloads of the syft package, which is PySyft itself rather than a client for a separate service. GitHub stars are also recorded, but only as corroboration, since a star is not an install.

Capability

4 medium confidence
4.0

PySyft runs computations across organizations without moving the data: each owner approves the collaborator and the job, the job runs on the owner's side, and an enclave package adds trusted-hardware execution. That matches Flower's reach, though the work travels over shared cloud storage rather than through a federation server.

  • https://cdn.jsdelivr.net/gh/OpenMined/PySyft@dev/packages/syft-enclave/README.md recorded 2026-09-27

    syft-enclave: "Enclave support for syft, enabling secure computation in Trusted Execution Environments (TEEs)", deployed with Docker and Terraform; the listed data owners' approval "gates every job on this enclave".

  • https://pypi.org/pypi/syft/json recorded 2026-09-27

    PyPI `syft` 0.10.0 (uploaded 2026-08-26): "PySyft : privacy-preserving data science over file-based transports"; "PySyft lets data scientists submit computations which are run by data owners on private data"; "Peer-to-peer with explicit auth — Data owners must approve each collaborator before any data flows"; "Isolated job execution".

  • https://ungh.cc/repos/OpenMined/PySyft/files/dev recorded 2026-09-27

    The 726-file tree at `dev`: packages/syft-enclave, packages/syft-bg, packages/syft-permissions and docs/auth.md beside the syft sync engine.

Verified 2026-09-27, except some scores, still awaiting re-confirmation