# AI4Privacy

> AI4Privacy builds open-source tools and datasets to protect personal data in AI systems. Grounded in the belief that privacy is a fundamental human right, it specialises in detecting, masking, and anonymising PII (Personally Identifiable Information), and publishes the world's largest open multilingual PII masking datasets for training privacy-preserving NLP models.

AI4Privacy's work spans four products: large-scale open training datasets, a privacy-preserving AI chat, a hosted PII detection API, and local SDKs for on-device detection. The open datasets are widely cited by research institutions and used by privacy teams worldwide, and are distributed under permissive licences via Hugging Face.

## Datasets

- [Datasets index](https://www.ai4privacy.com/datasets/): Overview of all open-source PII masking datasets for training and evaluating privacy-preserving models.
- [PII Masking 3M Asia-Pacific](https://www.ai4privacy.com/datasets/pii-masking-3m-asia-pacific/): The world's largest open multilingual PII masking corpus — 3M+ synthetic examples across 30 languages spanning Europe, the Americas, and Asia-Pacific. Latest release.
- [PII Masking 2M European](https://www.ai4privacy.com/datasets/pii-masking-2m-european/): 2M+ synthetic examples across 32 European locales and 98 entity types for privacy-preserving NLP.
- [OpenPII 1M](https://www.ai4privacy.com/datasets/pii-masking-1m/): 1.4M multilingual PII masking examples across 23 European languages and 19 entity types, open under CC-BY-4.0.
- [PII Masking 400K and below](https://www.ai4privacy.com/datasets/pii-masking-400k/): The foundational PII masking dataset series — 5 releases from 43K to 400K examples for training privacy-preserving NLP models.

## Products & API

- [Products overview](https://www.ai4privacy.com/products.html): Full product lineup — Open Datasets (build better models), AI Chat (protect user input), Privacy API (integrate fast), and Local SDKs.
- [AI Chat with Privacy](https://chat.ai4privacy.com/): Privacy-preserving chat that masks sensitive data before it reaches third-party models.
- [Privacy API](https://api.ai4privacy.com/): Hosted API for detecting and masking PII in text (access arranged via consultation).
- [Python SDK](https://pypi.org/project/ai4privacy/): Run PII detection locally, without sending data off-device.
- [JavaScript SDK](https://www.npmjs.com/package/@ai4privacy/ai4privacy): On-device PII detection for JavaScript/TypeScript projects.

## About

- [Home](https://www.ai4privacy.com/): Overview of AI4Privacy, its mission, products, and research citations.
- [Hugging Face — AI4Privacy](https://huggingface.co/ai4privacy): All open datasets, models, and demo spaces.
- [GitHub — AI4Privacy](https://github.com/AI4Privacy): Open-source code, SDKs, and tooling.

## Optional

- [LinkedIn](https://www.linkedin.com/company/ai4privacy/): Company updates and announcements.