AI transparency

Last updated: 2 October 2026

Manjo uses AI where it genuinely helps — turning messy sources into clean recipes — and is honest about it everywhere it happens. This page is our transparency disclosure under Article 50 of the EU AI Act, written for humans.

Our policy in one line: we mark what we generate, we surface what others mark, and recipe covers are always real food.

1. Where Manjo uses AI

Wherever you interact with an AI feature, the app says so at the point of use — you will never unknowingly converse with or rely on an AI system.

2. How AI content is labelled

3. What about images from elsewhere?

When you upload an image, we check it for existing C2PA content credentials before any processing, and images carrying AI credentials are labelled automatically. Honest limitation: an image without credentials is not proof it is real — not all AI tools mark their output, and metadata can be stripped. We surface what can be verified; we do not claim to detect what cannot.

4. Real photography for covers

Manjo never generates fake food photography. Recipe cover images are real photos — yours, or openly licensed photography with attribution. This is a product policy, not just a legal one: you should be able to trust that the dish on the cover exists.

5. AI can be wrong

Extraction and generation can make mistakes — quantities, units, ingredients, cooking times. Always sanity-check AI output, and take special care with allergies, dietary restrictions and food safety. AI suggestions are never dietary or medical advice.

6. Which systems we use

Text features run on Mistral AI models (EU); illustrations are generated on your iPhone by Apple's on-device image model, so the image is never generated on our servers. Providers may change as the product evolves — this page and the in-app labels are kept current. Questions: info@getmanjo.app.