# zama-ai/concrete-ml

Concrete ML: Privacy Preserving ML framework using Fully Homomorphic Encryption (FHE), built on top of Concrete, with bindings to traditional ML frameworks.

Repository: https://github.com/zama-ai/concrete-ml
Canonical: https://ross.abutalabs.com/products/concrete-ml
Language: Python
License: NOASSERTION
License Family: other
Topics: python, data-science, privacy, torch, scikit-learn, homomorphic-encryption, machine-learning, fhe, tfhe, ppml, fully-homomorphic-encryption
Last push: 2026-08-04T11:50:16+00:00

## Health v2 (maintenance only)
Score: 69/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 96, release rhythm 16, longevity 100
- inputs: {"age_days": 1624, "days_push": 29, "days_rel": 510, "gap_med": 95.5, "n_releases_24m": 3}
- flags: no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1445, forks 203 (observed 2026-08-28T04:04:45.165266+00:00)

## What it is
Concrete ML is a privacy-preserving machine learning library built on top of Zama's Concrete FHE compiler. It lets data scientists convert scikit-learn-style models and PyTorch neural networks into fully homomorphic encryption equivalents for inference or training on encrypted data without cryptography expertise.

## Use cases
- run machine learning inference on encrypted data
- train models without exposing sensitive training data
- analyze healthcare data while preserving patient privacy
- convert scikit-learn models to FHE equivalents
- deploy PyTorch models that operate on encrypted inputs
- build privacy-compliant ML services under strict data regulations

## When to choose
- you need ML on encrypted data without decrypting it
- your models use scikit-learn, XGBoost, or PyTorch APIs
- data privacy regulations prevent plaintext processing
- you want FHE without writing cryptography code

## When to avoid
- you need low-latency inference, as FHE adds large overhead
- your models rely on operations unsupported by FHE quantization
- you only need standard ML without privacy constraints
- you need a non-Python stack

## Facets
- artifact type: library
- maturity: active
- function: machine-learning, cryptography, privacy, security
- domain: machine-learning, privacy, data-science, security
- platform: python
- tags: fhe, fully-homomorphic-encryption, ppml, scikit-learn, pytorch, encrypted-inference, privacy-preserving-ml

## Member repositories
- zama-ai/concrete-ml (main) score 69

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:04:45.165266+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T04:36:15.857050+00:00, confidence not recorded.
  - readme: https://github.com/zama-ai/concrete-ml (fetched 2026-08-28T04:04:45.165266+00:00, sha 2922506b8169)
  - registry_pypi: https://pypi.org/pypi/concrete-ml/json (fetched 2026-08-29T11:46:13.375870+00:00, sha d775e8594b69)
- Data as of 2026-08-30T08:39:29.467469+00:00.