Language coverage

The supported set is the 24 official languages of the European Union plus Turkish and Azerbaijani. English is already an EU official language, so the total is 26.

What is measured

Every classifier is scored separately in all 26, at its calibrated threshold, on corpora generated per language rather than translated. There is no aggregate number, because an aggregate hides the tail and the tail is the point.

The PII model covers nine of the 26: en, ro, bg, hu, sl, hr, de, it and fr. The other seventeen are untested rather than unsupported, and closing that gap is a data task rather than a research one: each locale needs a national identifier generator with its real checksum, a phone country code, name lists and an email TLD.

Where it fails

Maltese is absent from the base model's pretraining. No amount of training data fixes that. It is the weakest language on every classifier and on two of them it scores zero. It stays in the published table, because a coverage table with the bad rows removed is not a coverage table.

Maltese and Azerbaijani national identifiers have no public checksum scheme, so those two can only be generated format-valid, which makes their labels weaker than the rest by construction.

What not to claim

The English locale in the training generator is labelled United Kingdom but uses a German identifier algorithm as a generic numeric fallback. A real UK National Insurance number carries no checksum, so a fallback is defensible, but do not state that English national identifiers are checksum validated.