What it detects

52 entity types. ✓ marks a checksum or format validator, so those recognizers confirm an identifier rather than matching its shape. Constants live in the entities package.

Group	Entity types
Generic	EMAIL_ADDRESS, PHONE_NUMBER, CREDIT_CARD✓, CRYPTO, IP_ADDRESS, URL, DATE_TIME, IBAN_CODE✓
United States	US_SSN✓, US_ITIN✓, US_PASSPORT, US_DRIVER_LICENSE, US_BANK_NUMBER, ABA_ROUTING✓, MEDICAL_LICENSE
United Kingdom	UK_NHS✓, UK_NINO✓
China	CN_ID_CARD✓
Australia	AU_TFN✓, AU_ABN✓, AU_ACN✓, AU_MEDICARE✓
India	IN_AADHAAR✓, IN_PAN, IN_PASSPORT, IN_VEHICLE_REGISTRATION, IN_VOTER, IN_GSTIN
Italy	IT_FISCAL_CODE✓, IT_VAT_CODE✓, IT_IDENTITY_CARD, IT_DRIVER_LICENSE, IT_PASSPORT
Spain	ES_NIF✓, ES_NIE✓
Singapore	SG_FIN✓, SG_UEN
Brazil	BR_CPF✓, BR_CNPJ✓, BR_RG, BR_CNH✓, BR_PIS✓, BR_CNS✓, BR_TITULO_ELEITORAL✓, BR_RENAVAM✓, BR_CEP, BR_PLACA, BR_PIX_KEY
Other	PL_PESEL✓, KR_RRN✓, FI_PERSONAL_IDENTITY_CODE✓, TH_TNIN✓

Every built-in detects a language-independent structured identifier: an IBAN or a Thai national ID looks the same in any surrounding text. The complete set therefore stays active under whichever language you build an engine with. The language key exists for language-specific recognizers, such as the ner module's model-backed recognizer.

CN_ID_CARD covers the 18-character citizen identification number in GB 11643-1999: six address digits, an eight-digit calendar date, three sequence digits and a MOD 11-2 check character. It accepts a transcribed lowercase x, preserves byte offsets next to Chinese text, and never joins columns or lines. It does not cover legacy 15-digit IDs, verify issuance or ownership, or validate historical address-code assignments. Checksum-valid findings are not proof that an identifier belongs to a real person. Tests use Appendix A examples and generated fixtures, not customer data.

Not in this list

PERSON, LOCATION, NRP and free-text DATE_TIME need a statistical model rather than a regex. They come from the optional ner or pfilter modules; the core alone never emits them.