Text
Detected PII
Detected personal data will appear here.
Small AI demo
Multilingual PII detection
Spot The 40 kinds of personal data, in 11 domainsIdentity 7Person nameSSNNational IDPassportDriver's licenseDate of birthTax IDContact 5EmailPhoneAddressPostal codeIP addressFinancial 6Credit cardIBANBank accountSWIFT/BICCrypto walletAmountCredentials 6API keyPasswordPrivate keyJWTConnection stringLogin credentialsOnline 2UsernameURLDevice 3MAC addressIMEIDevice IDLocation 1GPS coordinatesHealthcare 4Medical recordConditionMedicationHealth plan IDOrganization 1Company nameSpecial category 4ReligionPoliticalOrientationHealth statusLegal 1Case number across The 16 languagesEnglishGermanFrenchSpanishPortugueseItalianPolishRussianChineseJapaneseKoreanArabicHindiIndonesianVietnameseThai and redact them, in one pass. A 350M-parameter model that runs on a basic CPU, with no GPU.
Demo only. Your text is not stored. The model runs on this Space's server, not in your browser. The text is sent there only to find the PII, and the app does not store it or send it to any other service.
Base model
350M parameters
CPU only
LFM2.5-Encoder-350M-PII-Detector
Made by Liquid AI, not by AI for Data. This demo uses it as its base model.
- Type
- Bidirectional LFM2.5 encoder with a token-classification head
- Size
- 354M parameters, run in f32 on a CPU
- Detects
- 40 PII types in 16 languages
- Decode
- The model's own hybrid regex layer adds checks for formats such as email, IBAN, card numbers and JWT
- License
- LFM Open License v1.0