AI Transparency & Copyright
Farbwende – as of 3 November 2026
This notice supplements the imprint and the privacy policy and does not constitute binding legal advice. The German version is authoritative.
1. Scope
The following applies to all digital works that I publish or sell under the name Farbwende and that were created wholly or partly with generative AI. Three kinds of work are covered:
Images – created with the AI platform Midjourney
Audio and music – created with the AI platform Suno
Video and animation – animated clips from Midjourney, edited together in Final Cut Pro
It covers content on my website, on social media channels (such as Instagram @farbwende, YouTube), on streaming platforms (such as Spotify, Apple Music) and all individual commissioned works.
2. Workflow and internal documentation
Depending on the kind of work, my pieces come together in several steps. For every work I record the platform used, the model version, the prompt and all parameters, as well as the creation date. For Suno I additionally document that the work was generated during an active paid subscription and obtained through an official Suno download (“permitted download”), because since the Suno terms of service of 3 September 2026 the commercial usage rights depend on this (see section 4).
Images
1. AI generation – The source image is generated with Midjourney.
2. Upscaling – I then enlarge the image with Topaz Gigapixel AI and link the upscaled file internally to the original.
3. Photoshop finishing – For plain social media posts I limit myself to light colour or contrast corrections that do not reach the threshold of originality. For commissioned works I carry out extensive compositing, retouching and colour grading, so that an independent creative contribution arises. Every editing step is recorded in the associated PSD file and in a change log.
Audio and music
1. AI generation – The source audio track is generated with Suno.
2. Finishing in the DAW – Depending on the project, I either take the Suno output largely unchanged (only technical adjustments such as loudness or cuts that do not reach the threshold of originality) or I finish it comprehensively in my DAW (arrangement, sound design, mixing, mastering, my own recordings), so that an independent creative contribution arises.
Video and animation
1. AI generation – Individual animated clips are generated with Midjourney.
2. Composition and editing – The clips are selected, arranged, cut and combined with sound into an audiovisual work in Final Cut Pro. This dramaturgical arrangement and the editing are my own creative contribution.
3. Labeling and metadata
I label my AI-assisted works voluntarily and transparently:
Images carry the IPTC field “Digital Source Type: trainedAlgorithmicMedia”. On social media posts I add a visible note such as “AI image / Midjourney”. For commissioned works the note reads “AI image / Midjourney – editing: Farbwende” and the IPTC metadata additionally carry “© Farbwende [year]”.
Audio and music are labeled as AI-assisted wherever the respective platform provides for it (for example via the AI disclosure fields of distribution and streaming platforms).
Video and animation receive a visible note on their AI-assisted creation.
A C2PA manifest is embedded as a rule for commissioned works and, for now, on a trial basis for social media posts.
Note on the EU AI Act: The transparency obligations under Article 50 have applied since 2 August 2026. The machine-readable marking obligation (Art. 50 (2)) is primarily aimed at the providers of the AI systems (Midjourney, Suno), not at me as a user. My own obligation as a deployer mainly concerns the labeling of so-called deepfakes and certain content of public interest (Art. 50 (4)). For evidently artistic, creative, satirical or fictional works this obligation is limited to appropriate labeling that does not hamper the enjoyment of the work. My AI-generated works fall into this category and show no real persons; my voluntary labeling goes beyond what the law requires.
4. Copyright, usage and licensing rights
The legal situation differs by platform and by the extent of my own editing.
Images (Midjourney)
Social media posts: Since only minimal changes are made here, under current EU law no copyright of my own regularly arises. Third parties receive only a simple, non-exclusive licence of use, which is subject to the Midjourney terms of service.
Commissioned works: Through my extensive editing an independent creative contribution arises, so that I, Farbwende, hold a right of my own in it. Rights of use (simple or exclusive) are granted individually by contract.
Midjourney Inc. retains a worldwide, non-exclusive licence to all AI portions.
Audio and music (Suno)
I use Suno on a paid plan (Pro/Premier). Under the Suno terms of service as amended on 3 September 2026, Suno transfers to me the rights in outputs that I generated during an active paid subscription and obtained through an official Suno download (“permitted download”), and grants a commercial licence of use in them. Please note:
Suno transfers rights only to the extent that Suno holds them itself, and makes no promise that any copyright arises in an AI output at all. Commercial use no longer depends solely on the subscription being active at the time of creation, but on the official download via Suno; for works I obtained this way, the usage rights remain in place. Suno retains a worldwide, non-exclusive, sublicensable licence to the outputs and provides no indemnity against third-party claims.
For my own legal position the same logic applies as for images: if I take a Suno output largely unchanged, no copyright of my own regularly arises for lack of an own creative contribution. If I finish the output comprehensively in my DAW, arrangement, sound design, mixing and my own recordings create an independent creative contribution in which I hold a right of my own.
Video and animation
The individual clips generated with Midjourney are subject to the Midjourney terms of service (see above). The selection, arrangement, editing and scoring of the clips into an overall audiovisual work in Final Cut Pro are my own creative contribution, in which I hold a right of my own.
5. Data protection (GDPR)
My AI-generated works show no real, identifiable persons; I do not clone or imitate real voices, faces or the identity of specific real artists. The portrait photos on this website (for example on the “About” and “Contact” pages) are not AI works but real photographs by Sebastian Schüler, Frankfurt, and are labeled as such. No personal data whatsoever is processed in my AI works. Prompts contain neither brand names nor personal data.
Insofar as the operation of the AI platforms used involves a transfer of personal data to the USA, this is based primarily on the EU-US Data Privacy Framework (adequacy decision of the EU Commission), provided the respective provider is certified; in addition, the EU standard contractual clauses come into consideration as a fallback. Should real portraits or real voices ever be generated with AI in the future, this will be done exclusively with the express consent of the persons concerned.
6. Liability and warranty
Despite careful checking I cannot guarantee that no third-party rights are infringed in the AI training material of the platforms used. The providers (Midjourney, Suno) provide me with no indemnity against third-party claims. I address this residual risk through a consistently cautious way of working: no brand names, no protected characters and no imitation of specific real works, voices or styles in my prompts.
For commissioned works the liability and indemnity clauses agreed in each case apply.
I continuously follow the relevant legal developments – in particular the application of the EU AI Act, whose transparency obligations have applied since 2 August 2026, and the reform of product liability law (Directive (EU) 2024/2853, national implementation by 9 December 2026, which will also cover software and AI) – and adapt this statement as needed.
7. Updates
This notice is reviewed regularly and adapted as soon as legal or technical changes require it. Last updated: 3 November 2026.
← Back to home