---
title: "Who owns AI-generated design? What companies actually get when they buy it | minddraft"
description: "AI can generate images, logos and interfaces. Whether a company receives robust exclusive rights in them is a different question. A deep dive for Switzerland, Germany and Austria."
image: "https://minddraft.com/assets/img/insights/6/cover.webp"
---

[Insights](/en/insights) 

# Who owns AI-generated design? What companies actually get when they buy it

Date

08/25/2026

Author

![Marco Schär](/_ipx/q_80&s_48x48/assets/img/team/mschaer.webp) Marco Schär

read time

11 min

Actions 

![Open design source file with traceable human authorship beside an unattributed AI output](/_ipx/q_80&s_3072x1728/assets/img/insights/6/cover.webp)

Contents About 11 min left 

0% 

About 11 min left

[AI](/en/insights?filter=ai)

A company commissions a logo, a digital product or a new visual language. At the end, it receives files, screens and perhaps a broad assurance about the rights. But what was actually created, who is the author, and which rights can the supplier genuinely pass on? Generative AI turns these questions into practical concerns for procurement, contract review and long-term product ownership.

The short answer is: **AI can generate an outcome. It does not automatically generate the rights a company believes it is buying.**

This article examines the current position in Switzerland, Germany and Austria. It is not legal advice for a specific contract, registration or dispute. The analysis is current as of 25 August 2026.

## A file is not yet a rights position

Several different things are often merged in a project: the paid service, the delivered file, ownership of a physical copy, permission to use the work and a statutory exclusive right. They are not the same thing.

| Layer                  | What the company receives                                                        | Critical question                                        |
| ---------------------- | -------------------------------------------------------------------------------- | -------------------------------------------------------- |
| Service                | Research, concepts, selection, design and implementation                         | Was the agreed service performed?                        |
| Files                  | Figma, SVG, images, code, documentation and exports                              | Are the sources complete and editable?                   |
| Contractual permission | The use permitted between supplier and client                                    | Is the supplier entitled to grant that permission?       |
| Statutory rights       | Protection against third parties through copyright, trade marks or design rights | Did a right arise or was it registered?                  |
| Exclusivity            | The ability to exclude others from a use                                         | Does it bind only the supplier or third parties as well? |

Even an AI output without copyright protection can be delivered as a file and paid for as part of a service. The problem begins when the client expects an exclusive right enforceable against everyone, although that right may never have arisen.

## AI-generated is not the same as AI-assisted

The phrase “made with AI” is too vague for a legal or qualitative assessment. What matters is the task performed by the system and the decisions still made by a person.

| Use                           | Example                                                                   | Relevance to the final work                                                                |
| ----------------------------- | ------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------ |
| Research and idea development | Collecting perspectives, challenging assumptions, structuring information | AI influences the thinking process but is not the author of the final design.              |
| Internal prototype            | Making a product idea testable quickly                                    | The prototype is a decision tool and not necessarily the later client product.             |
| Assisted design               | Analysing original drafts, comparing options, automating repetitive work  | Protection depends on whether human decisions shape the specific result.                   |
| Generated final asset         | A logo, illustration or key visual taken directly from a model            | Human authorship and exclusive rights may be absent or difficult to prove.                 |
| Assisted code                 | Suggestions, refactoring, tests or repetitive implementation              | Licensing, provenance, security and human review remain part of the software supply chain. |

The useful distinction is therefore not between projects with and without AI. It is between a tool that supports human work and a system that is left to make the defining decisions about the final work.

## The legal starting point across the DACH region

The three legal systems use different terms but share the same foundation: copyright attaches to human intellectual creation. Swiss law expressly defines the author as the natural person who created the work. The Swiss Federal Institute of Intellectual Property likewise explains that AI output may be protected when a person contributes their own creative work and uses the model merely as a tool.

| Country     | Requirement                                                                        | Rights in contracts                                                                                         |
| ----------- | ---------------------------------------------------------------------------------- | ----------------------------------------------------------------------------------------------------------- |
| Switzerland | An intellectual creation with individual character; the author is a natural person | Copyright is generally assignable. Individual rights must be agreed precisely.                              |
| Germany     | A personal intellectual creation                                                   | Copyright itself is generally not assignable. Non-exclusive or exclusive rights of use are granted instead. |
| Austria     | A distinctive intellectual creation                                                | The author may grant permission to use the work or an exclusive exploitation right.                         |

Those differences matter when drafting contracts. The shared consequence is simpler: **where no protectable human work has been created, even the most precise contract cannot conjure up the corresponding copyright.**

This does not make an AI-generated asset automatically unusable or unlawful. Trade mark law, registered design rights, unfair competition law and contractual commitments may apply independently. Each has its own requirements. A logo may, for example, be registered as a trade mark, and a qualifying design may be registered. Neither route removes the need to examine provenance and earlier rights.

## What a 1,700-character prompt failed to achieve

The Local Court of Munich had to decide whether three logos created with generative AI were protected by copyright. In its judgment of 13 February 2026, case 142 C 9786/25, the court denied protection for all three. One logo was particularly revealing: the user had written and tested a prompt of around 1,700 characters.

For the court, effort, length and iteration were not decisive. Some instructions remained general or left the specific selection and design to the AI. Selecting one result from several proposals was not sufficient on its own. Copyright protection was possible in principle, but the human contribution had to shape the specific output objectively and identifiably.

The judgment is a German first-instance decision. It does not bind other German courts or courts in Switzerland and Austria. It nevertheless provides a useful practical measure: **prompting effort is not the same as design authorship.**

## The paradoxical risk: little protection of your own, yet potential infringement

A generated output may have no copyright protection of its own, or only uncertain protection, while still incorporating protected parts of someone else’s work. The Swiss Federal Institute of Intellectual Property notes that an output recognisably based on an existing work may qualify as an adaptation and may require a licence for its use.

This creates two separate assessments. Does the supplier hold a right it can pass on? And might the result infringe a third party’s rights? A positive answer to the first question does not resolve the second. Conversely, a lack of protection for the output itself does not make its use risk-free.

- **Copyright:**Are protected images, texts, illustrations or code elements recognisably reproduced?
- **Trade marks:**Does a sign conflict with an earlier mark or known identifier?
- **Design rights:**Is the design new and distinctive, or do earlier registered designs exist?
- **Personality rights:**Does the result depict identifiable people or deceptively similar representations?
- **Contracts and tool terms:**Which uses does the provider allow, and which guarantees are expressly excluded?

For sales and publications in the EU market, another layer has applied since 2 August 2026: the transparency obligations under Article 50 of the EU AI Act. They do not require every AI-generated design to carry a visible label. Providers of generative systems must generally mark synthetic outputs in a machine-readable format. Companies using those systems professionally must in particular disclose deepfakes clearly: AI-generated or manipulated image, audio or video content that resembles existing people, objects, places, organisations or events and could falsely appear authentic.

An evidently fictional or abstract T-shirt motif will not normally meet that definition. A photorealistic invented depiction of a real person or event may qualify as a deepfake. Article 50 does not generally require the disclosure to be printed directly into the motif. Where disclosure is required, however, it must be clear and perceivable by the time of first exposure. For evidently artistic or fictional works, it may be provided in a manner that does not hamper the display or enjoyment of the work.

## Why final design assets are particularly sensitive

An internal moodboard, a concept sketch and a final logo do not serve the same purpose. The more central and long-lived an asset is for a company, the more its provenance, exclusivity, editability and protectability matter.

- **Logos and symbols:**They identify a company over the long term and often need to be registered, defended and used across markets.
- **Key visuals and illustration systems:**They build recognition and evolve across channels for years.
- **Custom icons and UI components:**They become part of a product system maintained internally and extended into new features.
- **Characters and campaign figures:**They can develop into economically significant brand assets of their own.
- **Source files and design systems:**You need more than exports. You need traceable, editable and coherent working assets.

For assets like these, “we are allowed to give you the file” is a much smaller promise than “you receive a traceable, contractually robust rights position and can continue developing the work”.

## Code is different, but it is not free of rights questions

Software has long been assembled from many sources: proprietary code, open-source packages, frameworks, APIs, standards and generated artefacts. Open source does not mean “without rights”. It means that use is permitted under a particular licence. Licence management is therefore a normal part of professional software development even without AI.

AI assistance can create boilerplate, suggest tests, prepare refactoring or reduce repetitive work. Real developers must still review and take responsibility for architecture, security, dependencies, licences and the delivered code. A rapid internal prototype may also meet intentionally different standards from a production system. It simply needs to be treated clearly as a prototype.

The difference from design is therefore not a free pass. Software already has a composite and technically documentable supply chain. For a defining logo or key visual, however, the buyer often expects an individual, exclusively usable core asset with traceable human design authorship.

## What you should clarify before buying

Not every project needs a forensic dossier. If you are purchasing business-critical assets, however, you should examine more than price, aesthetics and delivery date.

1. **Ask for AI use to be disclosed:**Have the supplier show you where AI was used for research, exploration, prototyping or production.
2. **Define the final assets:**Clarify which parts are original human work, which come from third parties and which were generated.
3. **Review the chain of title:**Check whether rights from employees, freelancers and partners are contractually consolidated with the supplier.
4. **Document third-party material:**Have fonts, images, icons, libraries, models and the relevant licences documented transparently.
5. **Agree the source handover:**Record which sources you will receive: structured Figma files, vectors, components, tokens, code and documentation. Also clarify how decisions and intermediate states will remain traceable.
6. **Draft rights precisely:**Agree which assignable rights or exclusive rights of use you receive in scope, territory and time.
7. **Define the protection strategy:**Decide whether a logo should be registered as a trade mark, a design registered, or another asset given additional protection.
8. **Assess guarantees realistically:**Clarify which checks and commitments the supplier assumes. Tool terms and liability exclusions are not the same as a robust guarantee made to you.

These questions are not AI bureaucracy. They belong to a robust handover. Generative tools merely expose how little many projects have historically documented about provenance and rights.

## What this means when working with minddraft

When you work with minddraft, you do not receive a final design whose provenance becomes a black box after handover. You receive a reasoned, documented and evolvable system with clean source files. Third-party material such as fonts, images or libraries is not presented as original work but handled transparently.

AI helps us where it makes the work better: research, concept ideas, counterarguments, additional perspectives, internal prototypes, automation and development assistance. The goal is not to think less. It is to examine more and make better decisions in the same time.

A clear boundary applies to your final design assets: **what you receive as custom design is designed by our designers.** Logos, visual systems, illustrations, icons, interfaces and design systems are created as editable, structured source files under human responsibility. A generated internal prototype remains a tool for validation and does not quietly become your final product.

Responsibility remains clearly assigned during implementation as well. Our developers review architecture, code, dependencies and the production result. For you, that means AI may reduce repetitive work and provide suggestions, but it replaces neither professional responsibility nor a clean software supply chain.

## Conclusion: what matters is not whether AI appears in the process

The relevant question is not: “Was AI used somewhere?” It is: **what role did it play, who made the defining decisions, and what do you receive at the end?**

AI can broaden research, make ideas testable sooner and reduce tedious work. That is useful. Final assets intended to shape a brand or product for years also need human design, clean source files, documented third-party material and contracts that promise only what genuinely exists.

If you are investing heavily in a digital product or brand, you should therefore ask more than how the result looks. You should know how the work was created, how it can be edited and which rights you genuinely receive.

## How we put this responsibility into practice

Our follow-up article [“How we work with AI”](/en/insight/7/how-we-work-with-ai) explains our process: additional perspectives, early prototypes, human review and our own final design work. It also covers how we handle source files, usage rights and confidential information.

## Sources and legal notice

This analysis draws on the [Swiss Copyright Act](https://www.fedlex.admin.ch/eli/cc/1993/1798%5F1798%5F1798/en), guidance from the [Swiss Federal Institute of Intellectual Property](https://www.ige.ch/en/blog/blog-article/kuenstliche-intelligenz-koennen-ki-tools-urheberrecht-verletzen), the [German Copyright Act](https://www.gesetze-im-internet.de/englisch%5Furhg/englisch%5Furhg.html), the [Local Court of Munich judgment of 13 February 2026](https://www.gesetze-bayern.de/Content/Document/Y-300-Z-BECKRS-B-2026-N-1513) and the [Austrian Copyright Act](https://www.ris.bka.gv.at/GeltendeFassung.wxe?Abfrage=Bundesnormen&Gesetzesnummer=10001848). The Swiss Institute also explains the role of registered [design protection](https://www.ige.ch/en/protecting-your-ip/designs). The EU transparency analysis is based on [Article 50 of the EU AI Act](https://eur-lex.europa.eu/eli/reg/2024/1689/oj?locale=en) and the [European Commission implementation guidance](https://digital-strategy.ec.europa.eu/en/faqs/transparency-obligations-under-article-50-ai-act).

This Insight provides general orientation and does not constitute legal advice. Whether a specific work is protected, infringes third-party rights or can be assigned or licensed as intended must be assessed in the individual case.

## Want strong design without stumbling over sources and rights later?

Discuss your project 
[Call us ](tel:+41415620066)

![Marco Schär of minddraft](/_ipx/q_80&s_96x96/assets/img/team/mschaer.webp) 

Marco or someone from the team will get back to you, within 24 hours on working days.

## More insights.

[09/06/2026 6 min How we work with AI: more perspectives, better results AI UX](/en/insight/7/how-we-work-with-ai)

[06/03/2026 6 min Why we are building Wayfold: understanding what improves a website UX](/en/insight/9/why-we-are-building-wayfold)

[08/14/2025 8 min 8 design token mistakes and how to avoid them Design-Tokens Design-Systems](/en/insight/5/7-design-token-fails-that-can-throw-your-design-system-into-chaos-and-how-to-avoid-them)

[07/22/2025 4 min Accessibility is now mandatory, and remains a UX opportunity Accessibility UX](/en/insight/4/accessibility-is-now-mandatory-and-remains-a-ux-opportunity)

[06/17/2025 6 min UX that thinks ahead: Why proactive interfaces are the next SaaS growth lever UX AI UI](/en/insight/3/ux-that-thinks-ahead-why-proactive-interfaces-are-the-next-saas-growth-lever)

```json
{"@context":"https://schema.org","@graph":[{"@type":"WebPage","url":"https://minddraft.com/en/insight/6/who-owns-ai-generated-design-what-companies-actually-get-when-they-buy-it","inLanguage":"en","breadcrumb":{"@id":"https://minddraft.com/en/insight/6/who-owns-ai-generated-design-what-companies-actually-get-when-they-buy-it#breadcrumb"},"name":"Who owns AI-generated design? What companies actually get when they buy it | minddraft","description":"AI can generate images, logos and interfaces. Whether a company receives robust exclusive rights in them is a different question. A deep dive for Switzerland, Germany and Austria.","@id":"https://minddraft.com/en/insight/6/who-owns-ai-generated-design-what-companies-actually-get-when-they-buy-it#webpage","mainEntity":{"@id":"https://minddraft.com/en/insight/6/who-owns-ai-generated-design-what-companies-actually-get-when-they-buy-it#article"}},{"@type":"BreadcrumbList","@id":"https://minddraft.com/en/insight/6/who-owns-ai-generated-design-what-companies-actually-get-when-they-buy-it#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://minddraft.com/en/"},{"@type":"ListItem","position":2,"name":"Insight","item":"https://minddraft.com/en/insight/6/who-owns-ai-generated-design-what-companies-actually-get-when-they-buy-it"}]},{"@type":"Article","@id":"https://minddraft.com/en/insight/6/who-owns-ai-generated-design-what-companies-actually-get-when-they-buy-it#article","headline":"Who owns AI-generated design? What companies actually get when they buy it","description":"AI can generate images, logos and interfaces. Whether a company receives robust exclusive rights in them is a different question. A deep dive for Switzerland, Germany and Austria.","url":"https://minddraft.com/en/insight/6/who-owns-ai-generated-design-what-companies-actually-get-when-they-buy-it","mainEntityOfPage":{"@id":"https://minddraft.com/en/insight/6/who-owns-ai-generated-design-what-companies-actually-get-when-they-buy-it#webpage"},"image":"https://minddraft.com/assets/img/insights/6/cover.webp","datePublished":"2026-08-25","author":{"@type":"Person","@id":"https://minddraft.com/en/about#team-member-1","name":"Marco Schär"},"publisher":{"@id":"https://minddraft.com/#organization"},"inLanguage":"en","keywords":["ai"],"timeRequired":"PT11M","isAccessibleForFree":true}]}
```

```json
{"@context":"https://schema.org","@graph":[{"@id":"https://minddraft.com/#website","@type":"WebSite","description":"UX design for SaaS platforms, B2B applications and apps people enjoy using. Your digital product design lab in the Zug and Zurich region, Switzerland.","inLanguage":"en","name":"minddraft","url":"https://minddraft.com/","publisher":{"@id":"https://minddraft.com/#organization"}}]}
```