EU AI Act Notice
Last updated: 20 April 2026.
This document explains how Wearlo complies with Regulation (EU) 2024/1689 of the European Parliament and Council on artificial intelligence (the "EU AI Act") and what obligations apply to our customers (stores).
1. Wearlo's role
Wearlo provides an AI system for generating synthetic content — photoreal images of people in selected garments. Under the EU AI Act we take two roles:
- Provider — we supply the Wearlo product to fashion stores for integration into their product pages.
- Deployer — to the extent we showcase the product through demos on wearlo.ai using our own resources.
Our customers — stores offering the widget to their shoppers — act as deployers.
2. System classification
Wearlo is not a high-risk AI system under Annex III. It is a generative AI system subject to the transparency obligations of Article 50 — in particular the requirement to label synthetic content.
3. Transparency obligations (Article 50) — how we meet them
- Visible content labelling: every generated image carries a visible "AI" watermark baked into the image in contrasting colour. The shopper always knows she is looking at a visualisation, not a photo of an actual try-on.
- Image metadata: every result is tagged with EXIF/XMP metadata indicating synthetic origin (C2PA Content Credentials planned for Q3 2026).
- Pre-use disclosure: before opening the try-on, the shopper sees a short notice explaining that the result is AI-generated and a consent prompt (18+).
- No deepfake content: Wearlo does not reproduce the likeness of public figures or any person other than the shopper herself — the system prompt is closed and not editable per-tenant in v1.
4. AI model and its provenance
Generation uses the nano-banana-2 model provided by Kie.ai. It is a generative image-to-image model, trained on publicly available data per the provider's policy. Full information on the model's provenance and characteristics is available on request to supervisory authorities and enterprise customers.
5. Human oversight and risk management
- QA gallery: all generated images are reviewable in the admin panel. The Customer can flag a problematic generation in one click.
- Input filters: on upload we reject images that don't contain a person, are too low-resolution, or are flagged as potentially NSFW.
- Incident log: all reported generation issues are logged and reviewed by the Wearlo team within 48 hours.
- Prompt updates: prompt and model changes are versioned, and every change is preceded by testing against a reference set covering varied body shapes, skin tones, and garment types.
6. Store (Customer) obligations
The Wearlo Customer (store) is required to:
- Ensure shoppers know they are using a generative AI system (the Wearlo notice can be tuned to the brand's tone, but cannot be removed).
- Obtain freely given consent for the use of the shopper's likeness in AI generation.
- Not use Wearlo to generate images of anyone other than the shopper who personally uploaded the selfie (no generating third-party likenesses without consent).
- Not remove the visible "AI" watermark from generated images used in marketing. Only uses preserving the legible label are permitted.
7. Shopper rights
Shoppers have the right to be informed about AI use, the right to opt out (by not trying on), the right to deletion of generated images, and the rights granted by GDPR — described in the Privacy Policy.
8. Contact
Questions on EU AI Act compliance: [email protected].