Can an AI be listed as an inventor on a patent? Do AI-generated texts and images enjoy copyright protection? The DABUS decisions worldwide and German law provide clear answers – with far-reaching consequences for businesses.
When Algorithms Get Creative
An AI system designs a novel food container with fractal geometry. Another generates a photorealistic painting in the style of Old Masters within seconds. A large language model authors a technical article that is barely distinguishable from human work. The question arising from these developments is as fundamental as it is pressing: Who owns the intellectual fruits of artificial intelligence?
This question is no theoretical exercise. It concerns every company that uses AI for product development, content creation, or research. The answers that German and European law currently provide are clear – but not always what companies might wish for.
The DABUS Case: A Milestone in Patent Law
The most prominent test case worldwide was provided by the AI system DABUS (Device for the Autonomous Bootstrapping of Unified Sentience), developed by US researcher Stephen Thaler. DABUS reportedly autonomously produced two inventions: a food container with a fractal surface structure and a flashing beacon for emergency situations. Thaler filed patent applications for these inventions in numerous countries – naming DABUS as the inventor.
European Patent Office (EPO)
The European Patent Office rejected the applications as early as January 2020. The Board of Appeal confirmed the decision at the end of 2021: under Article 81 and Rule 19(1) of the European Patent Convention (EPC), the designated inventor must be a person with legal capacity. Since the right to the European patent vests in the inventor, the inventor must possess legal capacity. An AI has no legal personality – and "ownership of a machine" is not a valid legal basis for transferring inventor rights.
Federal Court of Justice (BGH) – X ZB 5/22
The German Federal Court of Justice (Bundesgerichtshof, BGH) clarified the German legal position with its decision of 11 June 2024 (case no. X ZB 5/22): an AI cannot be an "inventor" within the meaning of Section 37(1) of the German Patent Act (Patentgesetz, PatG). Only a natural person qualifies as an inventor.
However, the BGH opened a notable door: it allowed an auxiliary request in which the inventor designation read: "Stephen Thaler, who caused the artificial intelligence DABUS to generate the invention." This wording was deemed compatible with statutory requirements. The BGH emphasised that there are currently no AI systems that produce inventions entirely without human preparation or involvement. A human contribution that significantly influences the overall success of the invention suffices for inventorship.
International Context
The DABUS decisions present a uniform picture: in the USA, United Kingdom, New Zealand, and Australia (the latter after initial recognition by the Federal Court), DABUS was likewise not recognised as an inventor. WIPO accompanies the discussion at the international level through its conversation forum "Artificial Intelligence and Intellectual Property Policy", without having advocated for a change to the existing framework to date.
Section 37 PatG: Why Only Humans Can Invent
The German Patent Act (Patentgesetz, PatG) requires in Section 37 that the applicant designate the "inventor". The concept of inventor has always been understood by the legislature as referring to a natural person. The reasons are systematic in nature:
- Personality right: The right of inventorship is a highly personal right belonging only to natural persons
- Transferability: The right to the patent can be transferred – but this presupposes a right holder capable of legal acts
- Accountability: The patent system is based on the principle that a responsible human stands behind every invention
For practice, the BGH decision means: AI-generated inventions are fundamentally patentable – provided a human is designated as inventor who purposefully deployed the AI and significantly directed the inventive process.
Copyright Law: The Threshold of Personal Intellectual Creation
While patent law at least opens a path through human direction of AI, the situation in copyright law is more restrictive.
Section 2 UrhG: Personal Intellectual Creation
Under Section 2(2) of the German Copyright Act (Urheberrechtsgesetz, UrhG), only personal intellectual creations are protected by copyright. "Personal" here means: the work must have been created by a human being. Haufe explains that the concept of authorship in German law is inseparably linked to the natural person.
The consequence is unambiguous: Purely AI-generated content does not enjoy copyright protection. Someone who enters a prompt into a generative AI system and adopts the result unchanged is not the author of the generated work. The mere input of an idea – whether a text prompt or an image description – is insufficient, as the user makes no creative decisions of their own regarding the specific execution.
When Does Copyright Nevertheless Arise?
Copyright protection can arise when:
- Substantial human post-processing: The AI output is so significantly reworked by a human that an independent creative achievement exists
- AI as a tool: The AI is used merely as a tool while the human personally shapes the parameters that characterise the work – comparable to using a pen or a camera
- Selection and arrangement: The human makes a creative selection from a multitude of AI-generated drafts and arranges the overall work
The demarcation is fluid and will intensively occupy the courts in the coming years.
Text and Data Mining: The Copyright Side of AI Training
Another central aspect concerns the question of whether AI systems may be trained on copyright-protected works at all. The EU Copyright Directive 2019/790 created exceptions for this in Articles 3 and 4:
- Art. 3: Permits text and data mining (TDM) for scientific research
- Art. 4: Permits TDM for commercial purposes – but only if the right holder has not excluded the use through an opt-out (e.g., in machine-readable form)
Whether and to what extent training generative AI models falls under these exceptions is the subject of intense scholarly debate. For companies using AI-generated content, the question of liability for potential copyright infringements by the AI arises.
Practical Consequences for Businesses
The legal situation has immediate implications for companies using AI:
Adapting Patent Strategies
- Document the human contribution: When AI is used in your research and development process, carefully document which natural person directed the AI and significantly influenced the inventive process
- Formulate inventor designations strategically: The BGH decision shows that the wording of the inventor designation is crucial. Name the human as inventor and supplement the AI use as a tool
- Consider international filing strategy: The legal situation may develop differently across jurisdictions
Managing Copyright Risks
- No protection for pure AI outputs: Content generated exclusively by AI is in the public domain. Competitors may use it freely
- Consider trade secrets: Where copyright does not apply, protection as a trade secret (Geschäftsgeheimnisgesetz, GeschGehG) can represent an alternative – provided appropriate confidentiality measures are taken
- Liability for AI-generated content: Check whether AI-generated texts or images could infringe existing third-party copyrights. Liability falls on the user, not the AI
Rethinking Content Strategy
- Observe labelling obligations: Particularly in the context of the EU AI Act, transparency obligations exist for AI-generated content
- Document hybrid workflows: When humans and AI work on content together, document the respective contributions – this can be decisive for the question of copyright protection
- Review AI provider licence terms: The terms of use of many AI services contain clauses on intellectual property that should be carefully examined
Looking Ahead: Do We Need an AI-Specific Right?
The current legal situation – no inventorship or authorship for AI – creates a protection gap. Investments in AI-supported research and creative work produce results that may not be subject to any legal protection. WIPO is discussing at the international level whether a new sui generis protection right for AI-generated outputs should be created – similar to database protection in the EU.
Until such a solution is found, companies remain dependent on existing instruments: careful documentation of human contributions, strategic use of trade secrets, and contractual safeguards.
Conclusion: The Human Remains Indispensable – Legally
The answer to the title question is clear from a legal perspective: No, a machine cannot invent – at least not in the legal sense. Both patent and copyright law presuppose a human creator. AI is a powerful tool, but not a legal subject.
For businesses, this means: those who wish to deploy AI productively must consciously design and document the human contribution in the innovation process. Only then can the results be legally secured.
At compleneo, we support you with all questions relating to AI and intellectual property – from patent applications through copyright assessments to designing your AI usage strategy. Get in touch with us.
