Generative artificial intelligence tools present tremendous opportunities for businesses but also raise legal uncertainties. How can companies mitigate risks, particularly regarding copyright compliance? Sabrina Konrad, legal expert at the Swiss Federal Institute of Intellectual Property, shares her insights.
Nearly half of Swiss small and medium-sized enterprises have begun integrating artificial intelligence (AI) into their operations, according to a study by insurer AXA and the Sotomo Institute. Beyond translation and text generation, more than a third of companies use these technologies to create advertising content, while nearly a quarter employ them to generate images for their websites or marketing materials. This approach not only reduces costs but also provides rapid access to a vast range of creative options. But what about copyright? A company cannot simply repurpose a celebrity’s image for advertising – such as making James Bond wear a Patek Philippe watch. To what extent can businesses use AI-generated images without legal concerns? Sabrina Konrad, Deputy Head of the Copyright Law Legal Service at the Swiss Federal Institute of Intellectual Property (IPI), explains.
What risks do businesses face when using AI-generated images or text?
Sabrina Konrad: Depending on the quality of the tool, how it operates, and the prompts (i.e., instructions or data sets provided to an AI system) used, AI may generate content that includes copyrighted elements. As a result, businesses could face civil and even criminal penalties.
It is advisable to check whether legal exceptions, such as the right to citation, apply and, if so, to credit the original source. When in doubt, companies can also conduct an online search or use plagiarism detection tools.
Who owns content created by platforms like Midjourney or DALL·E?
Konrad: It depends. Businesses must carefully review the terms of use of these platforms, as they may impose restrictions on how AI-generated outputs (i.e., creations produced by AI) can be used. For example, under Midjourney’s current rules, the generated images belong to the individual or company that created them and can be used without crediting the platform. However, there are exceptions. Companies with an annual revenue exceeding one million dollars must subscribe to a specific plan. Additionally, if an image is generated from a high-resolution conversion of a pre-existing image, its use may be restricted.
From a legal perspective, businesses should avoid using these contents if they infringe on third-party copyrights. They should also be aware that some platforms may reuse the generated content for AI training or other purposes, including advertising.
Are there legal differences between AI-generated content and human-created content?
Konrad: In Switzerland, only content created by a human is protected by copyright. AI-generated content does not automatically benefit from this protection. However, if the user significantly influences or makes substantial modifications to the output, they may be eligible for copyright, as in this case, the AI is merely used as a tool. That said, there is no clearly defined threshold – such as requiring 60% modification – to establish authorship. For now, each case must be evaluated individually.
In the event of a dispute, how might the courts rule today?
Konrad: To my knowledge, no case on this subject has yet been tried in Switzerland. However, rulings have already been issued in other countries, notably in the United States and Germany, regarding the use of protected content for AI training. If a case were brought before the Swiss courts, they would likely take into account not only Swiss law but also foreign case law when making their decision.
Are there any guidelines in Switzerland for the use of generative AI in businesses?
Konrad: No, but the Swiss Federal Council has published recommendations for the federal administration, which can serve as a reference. The European Union’s AI Act also provides a relevant framework for Swiss companies, particularly those offering AI-based products or services in the European market.
Is it mandatory to disclose when content has been generated by AI?
Konrad: There is no strict legal requirement, but some providers, such as OpenAI, impose attribution under certain conditions. For businesses, being transparent about AI-generated content can help build trust with customers and employees. It is also important to understand how AI is used internally to develop a clear and consistent strategy.
What advice would you give to SMEs navigating these challenges?
Konrad: It is essential to be well-informed before using these tools – starting with reading the terms of use to understand their implications for data and copyright. Companies should also ensure the legality of AI-generated content. Additionally, AI can reinforce biases or produce questionable material, so maintaining ethical and quality oversight is crucial. AI is a powerful tool, but it should be used with caution and thoughtful consideration.
