For years, AI models and AI systems, service providers and users have been using artificial intelligence on protected works and neighbouring-right subject matter. These uses are mostly unauthorised, unpaid and opaque. They have no legal basis. Copyright and neighbouring rights, otherwise well established, can now be enforced only to a limited extent in the AI era. Creators and content producers are no longer prepared to be used as suppliers of raw material – with the inevitable consequence that cultural output is additionally forced to compete with mass-produced AI output. Litigation abroad shows that time is running out and legal certainty is lacking.
Across numerous jurisdictions, legal and political processes are now producing licensing models to authorise and remunerate AI training (training and adaptation of AI models and systems), AI retrieval (access to databases and current information) and similar uses. Licensing creates an ecosystem for AI content: without uncontrolled growth, infringement and opacity. Licensing gives rights holders the option to withhold their works and protected subject matter from AI use, or to consent only to specific research, products and services.
In Switzerland, the Federal Council and the Swiss Federal Institute of Intellectual Property (IPI) will present draft legislation in 2026. It is based on the Gössi motion, which was referred to the Federal Council in an amended form.
The recommendations for the CopA revision in the AI era are:
1. Strengthen copyright and neighbouring rights. The existing scope of rights already covers most AI uses, but clarity and legal certainty are lacking. The Copyright Act must expressly provide that AI training, AI retrieval and all other AI uses fall within the protected sphere of holders of copyright and neighbouring rights.
2. Principle of rights-holder consent. Artificial intelligence is now one of the ways in which linguistic, visual, musical, other acoustic and audiovisual works are used and exploited. Rights holders are entitled to decide on such distribution and use, whether through individual licence agreements or tariffs of collective management organisations.
3. Principle of rights-holder participation. Anyone who creates value with protected works and neighbouring-right subject matter, for themselves or for others, must give creative people and other rights holders a share. As long as effective technical enforcement is not possible – as envisaged in the European Union – remuneration must be provided for authors and other holders of copyright and neighbouring rights.
4. No free-use privileges. Copyright exceptions without remuneration should be avoided. AI systems of all kinds must be treated like other actors in the content exploitation chain. They either acquire contractual licences or benefit from a collective licence – always with a right to remuneration.
5. Research privileges with a right to remuneration. Free text-and-data-mining and similar research exceptions should be avoided. AI developments may begin as research, but are often supported by technology companies and sooner or later become a product – competing with the cultural works and protected subject matter used for their development.
6. Enforcement. Rights in works and neighbouring-right subject matter, and claims to remuneration, must be enforceable, including against foreign services used in Switzerland. This also prevents discrimination against domestic AI development and avoids disadvantaging Switzerland as an innovation and AI location.
7. Neighbouring right for media, including AI. If the neighbouring right for media becomes part of the legislative revision, its introduction is to be welcomed – without delay. AI summaries as answers to Internet searches are now commonplace, and the law must also capture this form of AI use and retrieval in all its forms as interference with copyright and neighbouring rights.
8. Collective management organisations. Collective licences will be central to the legislative design of the AI revision. Collective management organisations have experience with such licences and are able to expand their services. If opt-out declarations should prove necessary, collective management organisations must be involved at the forefront of the legislative project.
Current developments also call for a position on parliamentary initiatives and motions that have been debated or at least submitted. Our political demands are:
Confirm copying remuneration. A proven collective-management licensing model concerns copying remuneration paid by schools and organisations and remuneration for storage media used in the private sphere. The foundations of Common Tariffs 8, 7 and 4, and the remuneration under them, should be strengthened. The current proposals – the Aeschi parliamentary initiative and the Nantermod motion – should therefore be rejected without further ado.
Extended collective licences instead of privileges. In education and the library sector, sometimes extensive statutory privileges are under discussion. Where remuneration is paid, the participation principle is at least respected – as in the Binder-Keller motion of June 2026. Extended collective licences would have suited the National Library; instead, an exception that does not fit into copyright law was adopted in another statute, the National Library Act.
In the second example, the collective management organisations, the book sector and the cultural associations did not manage to secure a coherent copyright policy. The National Library, which unquestionably deserves support and respect, was too difficult a case for that. Tech giants, by contrast, can expect less support. The objection to this situational search for solutions is simple: copyright should not depend on sympathy. Rights holders deserve protection and payment whenever their works and protected subject matter are used by someone who does more than consume them and instead uses and distributes content.