Cookie Settings

We use cookies to improve your experience, analyze traffic, and support our marketing activities. You can accept all, reject non-essential, or customize your choices. Cookie Policy and Privacy Policy.

All systems running well
Skip to main content

AI Transparency and Acceptable Use Policy — Vizera AI

This is a translation. The German version is legally binding.

Last updated: 14 November 2026 · Version 1.4

1. What Vizera is

Vizera supports the generation and editing of architectural, interior and real-estate images as well as image-to-video outputs. You choose the function, the inputs and the model; the output is generated or edited by an external AI model. Vizera is therefore the provider of an AI system within the meaning of Regulation (EU) 2024/1689 (AI Act) that builds on third-party models.

Vizera outputs may contain factual, spatial, textual or visual errors. They are design and communication aids, not verified construction planning, structural calculations, permit documents, cost calculations or an assurance of how a project will actually be built.

2. Model providers and use of data

Vizera uses AI models from external providers. Depending on the function, your inputs are transmitted to the provider of the model used.

Vizera does not use your images or projects to train its own generally offered models. A customer-specific model is trained only on the basis of a separate written order and is used exclusively for that customer.

The terms of the model providers are not uniform. For the Google models, which include the default model for image generation, and for the prompt assistant, the providers' terms exclude the use of your inputs to train their models. For the other selectable models and for editing functions that use a fixed model (such as object removal, editing of masked areas, or a substitute model when the selected model is unavailable), the provider or the developer of the model may use inputs and outputs to improve its own models, or we have not been given an assurance that this is excluded. We will tell you on request which models and providers this concerns (privacy@vizera.ai). Provider terms may change.

Personal or confidential content may only be processed with models for which training use is excluded. Choosing a model does not replace any legal basis or confidentiality clearance that is required; for business customers, § 6 of the Data Processing Agreement applies in addition.

3. Notice of AI interaction

Where you are talking to an AI system — the prompt assistant in the app and the support assistant "Arizo" — this is made clear in the first message (Art. 50(1) AI Act). On request, support hands over to a human.

4. Marking of AI-generated images

Art. 50(2) of the AI Act requires providers of generative systems to ensure that artificially generated or manipulated outputs are marked in a machine-readable format and are detectable as such. Art. 50 has in principle applied since 2 August 2026. The transition period until 2 December 2026 for systems already on offer before that date concerns only the machine-readable marking and detectability of those systems. New systems are not automatically covered by this period; the classification must be assessed for each system and each change. The disclosure of AI interactions is not postponed by it.

Current status: Vizera does not currently add its own machine-readable provenance marking to outputs. Some model providers apply their own markings to their outputs (invisible watermarks or provenance metadata); it has not been demonstrated that these are preserved throughout processing and download. Image processing can remove metadata; the providers' watermarks have not been tested. From 2 December 2026 at the latest, Vizera writes a machine-readable marking as AI-generated or AI-edited into every image and video file it delivers (the IPTC metadata field "Digital Source Type") and preserves the markings supplied by the model provider in its own processing as far as this is technically possible. Invisible watermarks come from the model providers; not every model supplies one. Metadata can be lost when the file is later edited or passed on. We report on the status here.

5. Obligations of users

Anyone who publishes AI-generated or AI-edited images, videos or texts must check, depending on the context, their own labelling, personality-rights, copyright, competition or professional rules. Anyone who publishes image, audio or video content that deceptively resembles real persons, places, objects or events (so-called deepfakes) must, where the AI Act and its deployer obligations apply, disclose under Art. 50(4) that the content has been artificially generated or manipulated. A photorealistic depiction of an existing building — for example in renovation or virtual staging — may fall within this.

The exception for evidently artistic, creative, satirical or fictional works does not remove the obligation; it merely permits disclosure in a form that does not impair the presentation. Vizera does not decide whether a specific project falls under an exception.

For public use we recommend, irrespective of the state of technical implementation, a visible notice, for example: "Created or edited with AI using Vizera AI."

6. Prohibited use

The following in particular are not permitted:

  • non-consensual intimate content or sexual content involving minors;
  • deceptive impersonation, fraud, phishing or unlawful deepfakes;
  • discriminatory or unlawful content, or content glorifying violence;
  • infringement of data protection, copyright, trade marks, trade secrets or personality rights;
  • circumvention of protective mechanisms, limits or access controls; and
  • use of an output as the sole basis for safety-critical or legally regulated decisions.

7. Moderation and remedies

Vizera may restrict inputs, outputs or accounts where there is a substantiated violation. Those affected receive a statement of reasons setting out the applicable rule, the facts relied on, any use of automated means and the possibility of complaint. Presumably illegal content can be reported through the reporting procedure described in the legal notice (Impressum).

Version 1.4 · 2026-11-14