Whether copyright subsists or not is decisive, and the principle is the same across all categories of works:
- The author can claim the right to be named; performers have their own personality-related rights.
- Rights holders can prohibit others from making certain uses of the work; these are exclusive rights.
- Rights holders can grant licences and demand remuneration; these are economic rights.
- Collective management organisations collect royalties for certain uses and distribute them to holders of copyright and neighbouring rights.
For authorship, the law provides a presumption in favour of the person named. Publishers and producers can assert their own or derived rights and make that claim visible through the © symbol and other notices.
The more output is generated by machines, and the greater the machine contribution to human creations, the more useful it becomes if such information can not only be asserted but also proved.
A wide range of AI markings, provenance evidence, timestamps, identity checks, detection programs and other instruments are now emerging to help distinguish machine-made from human works. They do not replace copyright, and the availability of such tools should not be used against holders of copyright and neighbouring rights. Technical evidence is neither necessary nor sensible in every case. For important documents and files, however, it can be useful to make the protection status recognisable.
The Authors Guild in the United States launched the «Human Authored» label in 2025; the Society of Authors in the United Kingdom followed in 2026 with the same designation. Writers can use it to declare that a work was essentially written by humans.
Since 2 August 2026, the transparency obligations of the EU AI Act have also applied: providers of generative AI must mark synthetic content in a machine-readable form; deepfakes and certain AI-generated texts on matters of public interest must be visibly disclosed where there is no human review or editorial responsibility.
ProLitteris is joining this small tradition, but we are going two steps further. With the technological experience of the IT company Swiss Trust Layer in areas where trust and evidential value are particularly important, we are designing a Swiss infrastructure that strengthens transparency in the production of creative content and the rights connected with it. We are currently inviting further collective management organisations and companies to use the service and become partners. The plan is for the new evidence infrastructure to meet further requirements for digital assets and also support legal transactions with AI users.
MyCopyright builds on the voluntary icons published by the European Commission. We add the self-declaration on AI use (yes, no, partly) to the digital fingerprint of the file. And we distinguish four levels of AI use: AI GENERATED, AI MODIFIED, AI ASSISTED and AI FREE. HUMAN AUTHORED and HUMAN APPROVED are added – these two labels require qualified identification of at least one natural person, for authorship and for responsibility, two roles that ultimately have to be assumed by humans.
The MyCopyright labels are defined as follows.
AI GENERATED
AI-determined content.
The content was generated entirely, predominantly or in its defining elements by AI.
Based on a prompt or chat, AI generates a complete article, an illustration, a music or voice recording, or a video clip. Predominantly AI-generated content with individual human corrections also falls into this category.
AI MODIFIED
Existing content modified by AI.
The content is based on existing content and was modified by AI in at least one defining element.
AI modifies or adds at least one defining element in existing content: for example the statement of a text, a person in a photograph, the spoken words in an audio recording or the sequence shown in a video.
AI ASSISTED
Human-determined content with an AI contribution.
The content was predominantly generated by humans, but at least one defining element was generated or modified by AI.
Humans create the content, but AI contributes at least one defining element: for example the headline of an article, the background of an image, an additional voice in a recording or a generated video sequence.
AI FREE = HUMAN AUTHORED
Human content without an AI contribution. The definitions of AI FREE and HUMAN AUTHORED are the same, but the conditions for use differ.
The content was generated entirely by humans. Minimal technical, preparatory or checking use of AI without effect on a defining element is disregarded.
Examples are a text with automated spell-checking, a photograph with a simple light correction, a music or voice recording with noise reduction, or an unchanged smartphone video.
HUMAN APPROVED
Content reviewed and taken responsibility for by humans.
The complete content was reviewed and approved by at least one identified human. That person and/or a designated organisation assumes responsibility for it.
An identified person reviews and finalises complete content before publication: for example a manuscript, a customer letter, a media release, a social-media post, an advertising graphic or an information video.
Rules for use and combination
ProLitteris and Swiss Trust Layer offer the HUMAN AUTHORED and HUMAN APPROVED labels only to persons who have completed qualified identification. Identification takes place in a technical KYC process using the passport document and the technical features it contains. The solution is provided by Swisscom Trust Services, as are the trust services for sealing documents by means of a fingerprint and a qualified timestamp.
Exactly one of the labels AI FREE, AI ASSISTED, AI MODIFIED or AI GENERATED applies to each item of content. HUMAN AUTHORED can be combined only with AI FREE; HUMAN APPROVED can be combined with any of these labels.
Where categories overlap, the order is AI GENERATED, AI MODIFIED, AI ASSISTED, AI FREE. In cases of doubt, the label revealing the greater degree of AI use must be selected.
A label always refers to a specific version of the content. A new edit can lead to a different label; a later reclassification is documented with the date and a correction notice.
For composite content, one label applies to the content as a whole. Defining components created in a different manner are additionally identified separately in the declaration.
Alongside the labels, ProLitteris is addressing a second area with MyCopyright: rights holders’ opt-out declarations towards AI models. The EU Copyright Directive allows rights holders to exclude certain text-and-data-mining uses by means of a machine-readable reservation of rights. In practice, the system is incomplete: its effectiveness and enforcement depend on technical standards, the discoverability of the reservation and compliance by providers. Our solution again starts with evidence: digital fingerprints, watermarks and metadata are deposited in a traceable manner and independently managed by ProLitteris. Qualified trust services secure the timing and integrity of the declaration.
MyCopyright is not the only initiative addressing several copyright problems around artificial intelligence by technical means. We monitor other ideas and developments and are ready to assess such solutions and, where suitable, connect them to our platform. The name «MyCopyright» will sooner or later be supplemented by «Swiss Copyright Office». The important point is that we do not pre-empt any legislative solution and continue to support stronger copyright, where appropriate with collective licences for the use of artificial intelligence.
We hope that this will soon make it crystal clear again whether a beautiful poster or moving music was created with human creativity and/or machine skill – and, more importantly, whether it contains rights and assets belonging to humans. Trust is a valuable asset, including for creative content with both human and machine-generated elements.