A verdict, not a score.

humaniser.eu is an EU-resident AI-text detector and editorial humaniser operated by Ariada (Agonist Development AB, Sweden). It returns a verdict — human, AI, or manual review — rather than a single percentage, with a signed audit receipt and per-locale false-positive rates published per language.

All inference runs on infrastructure inside the EU. Submitted passages are not retained after the verdict is returned unless you opt into receipt persistence. Logs containing text content are kept only for the duration required to produce the signed receipt, then purged. No transatlantic transfer.

Each verdict ships with a signed JSON receipt — text hash, model SHA, calibration card reference, timestamp — designed as decision-aid for an academic-integrity panel, an editor, or an HR reviewer. It is not a solely automated decision under GDPR Article 22; a human reviewer must make the final call.

False-positive rates are published per language on the calibration page. English text scored on the v0.3 model holds AUROC near 0.94 at FPR 5% on the RAID-style benchmark; rates for Mandarin-L1 and lower-resource European languages run higher and are disclosed honestly, not averaged away.

Originality.ai publishes a percentage score and English-centric accuracy claims. humaniser.eu publishes per-language calibration cards, refuses to score when the model abstains, hosts entirely in the EU, and ships every verdict with a signed Merkle-anchored receipt. The two are not interchangeable for EU regulated workflows.