# openai/privacy-filter

OpenAI Privacy Filter

Repository: https://github.com/openai/privacy-filter
Canonical: https://ross.abutalabs.com/products/privacy-filter
Language: Python
License: Apache-2.0
License Family: permissive
Last push: 2026-04-22T19:55:02+00:00

## Health v2 (maintenance only)
Score: 49/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 78, release rhythm 35, longevity 9
- inputs: {"age_days": 138, "days_push": 133, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases, young
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 2670, forks 239 (observed 2026-08-28T04:07:09.380120+00:00)

## What it is
OpenAI Privacy Filter is a small bidirectional token-classification model (1.5B total, 50M active parameters) for detecting and masking personally identifiable information in text. The repo provides a Python package and CLI (`opf`) for running one-shot redaction, evaluation, and fine-tuning of the model on-premises.

## Use cases
- redact PII from text files before sharing or training
- sanitize logs and datasets on-premises
- detect names, dates, and other PII in long documents
- fine-tune a PII detector on custom data distributions
- run PII masking locally in a browser or on a laptop
- pipe text through a privacy filter in shell one-liners

## When to choose
- you need fast, context-aware PII redaction that runs locally without sending data to external services
- you need a permissively licensed, fine-tunable PII model for commercial deployment
- you process long documents (up to 128k tokens) without chunking
- you want configurable precision/recall tradeoffs for redaction

## When to avoid
- you need structured document (PDF/image) PII redaction rather than plain text
- you require certified compliance guarantees rather than best-effort ML detection
- you need a managed cloud PII API with vendor support
- you lack the hardware to run even a small transformer model

## Facets
- artifact type: library
- maturity: active
- function: nlp, machine-learning, cli, privacy, security
- domain: privacy, machine-learning, security, developer-tools
- platform: python, cli, cross-platform
- tags: pii-detection, pii-masking, data-sanitization, token-classification, redaction, on-premises, fine-tunable, viterbi-decoding, natural-language-processing, gpu, cpu

## Member repositories
- openai/privacy-filter (main) score 49

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:07:09.380120+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 2026-08-30T02:18:11.293107+00:00, confidence not recorded.
  - readme: https://github.com/openai/privacy-filter (fetched 2026-08-28T04:07:09.380120+00:00, sha 79c223290f1a)
- Data as of 2026-08-30T08:39:29.467469+00:00.