A technology company releases a generative AI product trained on thousands of copyrighted works. The tool produces original-looking images, marketing copy, and music. The company's commercial strategy depends on owning the outputs. Then counsel delivers the news: under German intellectual property law, the question of who – if anyone – owns an AI-generated work is neither settled nor straightforward.
AI-generated works in Germany occupy a doctrinal grey zone. German copyright law ties protection to a human author's personal intellectual creation, meaning fully autonomous AI outputs receive no copyright protection by default. The ownership question turns on how much creative direction a human contributor exercised and how that contribution can be documented.
This analysis covers the doctrinal foundations of copyright protection in Germany, competing interpretations emerging from courts and academic practice. The gap between statutory text and commercial reality, cross-border implications for European clients. Additionally, a practical strategic outlook for businesses building on or with generative AI.
Doctrinal foundations: why German copyright law and AI outputs clash
German intellectual property legislation rests on a principle that has served it well for over a century: copyright protection attaches to works created by a human being who exercises personal intellectual creativity. This standard is known as the requirement of personal intellectual creation – a threshold grounded in the civil law tradition that shapes the entire German approach to authorship.
The relevant branch of legislation does not define an author as a legal entity or a machine. Courts in Germany have consistently interpreted authorship as an inherently human concept. The Bundesgerichtshof (Federal Court of Justice of Germany) has reaffirmed in multiple contexts that a work must reflect an individual author's creative personality. An output produced without that human ingredient falls outside the statutory protection regime.
This creates an immediate structural problem for generative AI outputs. When a large language model, image generator. Alternatively, music synthesis tool produces a work in response to a prompt. The output reflects statistical patterns learned during training rather than the expressive choices of a human author. The question of whether the human who wrote the prompt, the developer who trained the model, or the company that deployed it has contributed enough personal intellectual creativity to qualify as an author remains open.
German courts have not yet delivered a definitive ruling on fully autonomous AI-generated works. What the case law does supply is a well-developed body of doctrine on computer-assisted works – cases where a human uses software as a tool but retains creative control. Under that doctrine, protection is available when the human's creative decisions dominate the production process. The software is treated as a sophisticated brush. The emerging question is where generative AI sits on the spectrum between brush and independent creator.
A further complication arises from the distinction between the level of creativity required for different categories of protected subject matter. Photographs, for example, historically attracted a lower threshold in German intellectual property legislation, even before full harmonisation with EU standards. Software itself is protected as a literary work. Applied art works require a degree of creative height that courts have assessed case by case. Each of these categories interacts differently with AI authorship questions, and the doctrinal answer may vary depending on the type of output involved.
Competing interpretations and the gap between statute and practice
German intellectual property scholars and practitioners are divided across three broad positions. Understanding each is essential for international clients designing IP strategies around AI-generated content.
The first position – strict formalism – holds that the personal intellectual creation requirement is non-negotiable. AI outputs receive no copyright protection regardless of how elaborate or commercially valuable they are. Under this view, AI-generated works enter the public domain the moment they are produced. Businesses that build product lines on these outputs cannot prevent competitors from copying them. The practical consequence is severe: an investment in generative AI content creation yields no proprietary asset.
The second position – contribution theory – focuses on the human creative input that precedes the AI's operation. Proponents argue that a sufficiently detailed and creative prompt, or a carefully curated selection of training data, constitutes a personal intellectual contribution capable of sustaining copyright in the final output. The author under this theory is the person who directed the AI's generative process. Courts in Germany have not formally adopted this position, but practitioners in IP-intensive sectors increasingly rely on it when advising clients on documentation strategies.
The third position – derivative protection theory – draws on the existing doctrine covering computer programs and databases. Some scholars argue that where an AI system produces outputs that can be characterised as a form of compiled or organised information. Protection may be available under the database right or the software protection provisions of German intellectual property legislation, even absent personal intellectual creation. This argument has practical appeal but faces doctrinal resistance: German courts have traditionally been careful to maintain the boundary between copyright protection (which rewards creativity) and related rights protection (which rewards investment and effort).
The gap between statute and practice is most visible in commercial transactions. When a GmbH (private limited liability company) acquires an AI-content platform, due diligence teams encounter IP representations that cannot be fully substantiated under current doctrine. Technology licensing agreements purport to transfer rights in AI outputs that may not legally exist. Employment contracts assign to the employer the IP created by AI tools used by employees – on the assumption that such IP exists and is assignable. Practitioners advising on these transactions must either negotiate around the uncertainty or insert contractual risk allocations that acknowledge it.
The Amtsgericht (local court) and intermediate appellate levels have not yet produced a consistent line of authority on the core question. The handful of decisions touching AI and copyright to date have mostly addressed adjacent issues: copyright infringement by AI training, the admissibility of AI-generated evidence, and the liability of platform operators. The doctrinal question of AI authorship is working its way up the judicial hierarchy. Until the Bundesgerichtshof or the European Court of Justice delivers a ruling that German courts must follow. The practitioner's answer to "does my company own this AI output?" remains "it depends. Additionally, the law is in flux."
For international clients, this uncertainty is compounded by the territorial nature of copyright. A work that attracts protection in the United Kingdom. where statutory provisions expressly contemplate computer-generated works and assign ownership to the person who made the arrangements for the creation of the work. may receive no protection in Germany. A cross-border content strategy built on a single IP ownership assumption is therefore legally fragile. See our analysis of AI-generated works and intellectual property in Portugal for a parallel treatment under Portuguese law, which shares the civil law authorship tradition but has developed its own doctrinal responses.
AI Act compliance and the interaction with IP law
Germany sits at the intersection of two converging regulatory developments: the EU AI Act and the evolving German approach to intellectual property in AI outputs. These two bodies of law address different problems, but they interact in ways that create both compliance obligations and strategic opportunities.
The EU AI Act introduces a risk-based system of AI Act compliance obligations. General-purpose AI models – including most of the large generative AI systems used to produce text, images, and code – face specific transparency and documentation requirements. Providers must summarise the copyrighted training data used to train these models. This transparency obligation sits in direct tension with the doctrinal uncertainty around AI-generated works: if copyright protection for AI outputs depends in part on documenting human creative contribution. The AI Act's disclosure requirements may either support or undermine those documentation strategies depending on what they reveal.
Algorithmic accountability is a related dimension. Under emerging EU and German regulatory thinking, deployers of AI systems bear responsibility for outputs that cause harm – including outputs that infringe third-party copyright. Software liability in Germany draws on general tort principles and, increasingly, on sector-specific digital services legislation. A company that deploys a generative AI tool for commercial content creation assumes liability exposure not only for the outputs it produces but for the training data that shaped those outputs. Where that training data included copyrighted works, the legal position on whether the AI provider or the deployer bears liability for downstream infringement is unresolved.
Technology licensing arrangements involving AI systems require particular care. A licence that grants rights to use an AI model to generate content does not, by itself, resolve the question of who owns that content or whether any copyright subsists in it. Practitioners in Germany are beginning to see technology licensing agreements that include express warranties from AI providers about training data provenance and indemnities for third-party copyright claims. These provisions are becoming a standard ask in negotiations – and their absence from existing agreements represents a significant gap for companies that concluded technology licensing deals before these issues crystallised.
The digital services regulatory environment also intersects with IP law through the liability regime for online platforms. A platform that hosts and distributes AI-generated content may face claims from original rights holders who assert that the content was derived from their works without authorisation. The doctrinal interaction between copyright infringement analysis and the automated generation process is contested. Some German courts have assessed the question by reference to the training process; others have focused on the similarity of the output to the allegedly infringed work. Neither approach has achieved dominance.
For practical guidance on how these AI Act and IP compliance obligations interact in the German market. Our team's dedicated service for AI and technology law in Germany covers the full range of regulatory and IP strategy questions arising from AI deployment.
To explore how the intersection of AI Act compliance and intellectual property protection applies to your business in Germany, contact us at info@ferrazwhitmore.com.
Cross-border implications for European clients
Germany is Europe's largest economy and a primary target market for technology businesses operating across the continent. The doctrinal gap between German IP law and the legal positions of other major European jurisdictions creates specific risks for international clients.
The EU harmonisation programme that produced the Information Society Directive and the Database Directive did not address AI authorship. Harmonisation in this area remains a project in progress. The result is that EU member states apply a common structural framework. personal intellectual creation as the copyright threshold. but reach divergent practical conclusions about what "personal" and "intellectual" mean when a human interacts with an AI system. German courts have historically applied a relatively demanding version of this standard. French and Italian approaches, while sharing the civil law foundation, have been more receptive in some contexts to expansive readings of creative contribution.
For a business operating across the EU, this divergence means that the IP ownership map for AI-generated content must be drawn jurisdiction by jurisdiction. Content produced by a tool deployed in Germany may be unprotected there while attracting protection in France. A licence of that content purporting to cover all EU territories may therefore transfer rights that do not exist in part of the territory.
The United Kingdom presents a specific cross-border challenge. Post-Brexit, English intellectual property legislation retains a provision that expressly protects computer-generated works and assigns authorship to the person who made the arrangements for the work's creation. This statutory mechanism has no German equivalent. A company structured to hold IP assets in the UK may achieve protection for AI outputs that a German-law structure cannot replicate. Cross-border IP holding strategies must account for this asymmetry. including the question of whether a UK-incorporated entity can enforce its rights in German courts when the underlying German law does not recognise the protected interest.
The Insolvenzordnung (German Insolvency Regulation) adds an insolvency law dimension for distressed businesses. Where a company holding AI-generated content becomes insolvent, the insolvency administrator's ability to realise value from that content depends on whether the content is legally protected. Unprotected AI outputs may form part of the insolvent estate in a different way from copyrighted works – affecting the administration strategy and the recoveries available to creditors. This is a non-obvious risk that surfaces in the context of AI-content businesses seeking investment or contemplating restructuring.
Registration in the Handelsregister (German Commercial Register) reflects a company's legal identity and, by extension, its capacity to hold IP rights in Germany. A foreign entity entering the German market to exploit AI-generated content should assess whether its corporate structure. including the choice between a branch office. A subsidiary GmbH. Alternatively, another vehicle. positions it to assert and defend whatever IP rights may be available. The Handelsregister filing is the foundational step for a corporate presence capable of entering into enforceable technology licensing arrangements under German commercial legislation.
For a detailed treatment of German intellectual property law outside the AI context. including trademark registration, enforcement mechanisms. Additionally. Licensing best practices. our comprehensive analysis of intellectual property law in Germany provides the background framework against which AI-specific questions arise.
For a tailored strategy on cross-border IP structuring and AI Act compliance in Germany, reach out to info@ferrazwhitmore.com.
Strategic recommendations and forward outlook
The doctrinal uncertainty around AI-generated works in Germany creates a clear strategic imperative: businesses cannot wait for legislative clarity before making IP allocation decisions. The competitive window for establishing strong contractual and documentation practices is open now. Organisations that build these practices will be better positioned when courts and legislators eventually resolve the remaining questions.
The following strategic framework reflects current best practice for international clients operating with generative AI tools in the German market.
Document human creative contribution at every stage. The strongest argument for copyright protection in AI-generated outputs rests on demonstrating meaningful human creative direction. Businesses should implement workflow documentation that captures prompt design decisions, iterative refinements, selection choices among multiple AI outputs, and editorial interventions applied to raw AI-generated material. This documentation serves two purposes: it strengthens the copyright claim and it provides evidence if the company's ownership is challenged.
Structure technology licensing agreements for uncertainty. AI tool licences should expressly address the ownership of outputs, include representations about training data provenance, and provide indemnities for third-party copyright claims arising from the AI's training process. Existing agreements that lack these provisions should be reviewed and, where possible, amended. The absence of training data representations is among the most common and consequential gaps practitioners encounter when auditing AI-related agreements.
Use contractual protection as a substitute for statutory protection. Where copyright protection is legally uncertain, trade secret law and contractual confidentiality obligations provide an alternative protective layer. Under German trade secret legislation, commercially valuable information that is subject to reasonable confidentiality measures qualifies for protection regardless of its creative origin. This does not replicate copyright – it cannot prevent independent creation of similar works – but it protects against misappropriation by persons with access to the AI outputs or the underlying models.
Monitor the AI Act compliance timeline. The EU AI Act's requirements for general-purpose AI model providers enter into force on a phased schedule. Businesses that develop or deploy such models in Germany must assess their obligations under the risk classification system and implement the required documentation, transparency, and testing measures. Failure to comply creates regulatory exposure that compounds the existing IP uncertainty.
Assess cross-border IP holding structure. Given the divergence between German and UK law on computer-generated works, businesses with significant AI content operations should evaluate whether a multi-jurisdiction IP holding structure is commercially justified. This involves weighing the cost and administrative complexity of maintaining IP assets in multiple jurisdictions against the value of securing protection in jurisdictions where German law does not extend it.
The outlook for legislative resolution is measured. The European Commission has indicated that the relationship between AI and copyright will require legislative attention, but no proposal for harmonising AI authorship rules is imminent. The Bundesgerichtshof is likely to address core questions of AI authorship within the next several years as cases work through the appellate system. In the meantime, German courts at intermediate levels may produce divergent rulings. The period of uncertainty is unlikely to end quickly, and the competitive consequences for businesses that fail to manage it proactively are significant.
A non-obvious strategic opportunity exists in this period of uncertainty. Businesses that invest now in documentation practices, contractual structures, and training data compliance will possess competitive advantages that go beyond legal protection. They will have cleaner IP assets for due diligence purposes, stronger negotiating positions in technology licensing transactions, and lower litigation exposure as enforcement activity around AI-generated content increases. The organisations that treat this period as a reason for inaction will face greater costs and greater risks when the law crystallises.
Frequently asked questions
Q: Can a company in Germany own copyright in works produced entirely by an AI system without human creative input?
A: Under current German intellectual property law, no. Copyright protection requires a personal intellectual creation by a human author. Works produced entirely by an AI without meaningful human creative direction fall outside the scope of statutory protection and enter the public domain. Businesses relying on AI-generated content for competitive advantage should therefore invest in documentation practices that support a human-contribution argument and in contractual mechanisms that provide alternative forms of protection.
Q: How long does it typically take for AI-related IP disputes in Germany to reach a definitive resolution through the courts?
A: IP disputes in Germany can take several years to reach the Bundesgerichtshof, which is the court whose decisions set binding precedent. Proceedings at first instance before the Landgericht (Regional Court) and subsequent appeal to the Oberlandesgericht (Higher Regional Court) typically take a combined period of two to four years for contested matters. A single question of AI authorship doctrine could take considerably longer to achieve final judicial settlement. Engaging a lawyer in Germany with cross-border IP experience early in a dispute is advisable given the complexity and duration of proceedings.
Q: Does the EU AI Act resolve the question of copyright ownership in AI-generated works?
A: No. The EU AI Act is a product safety and risk-management regulation, not an intellectual property instrument. It imposes obligations on AI system providers and deployers – including transparency requirements about training data and algorithmic accountability measures – but it does not assign or create copyright in AI outputs. A law firm in Germany advising on AI Act compliance will address a different set of obligations from those arising under intellectual property legislation. Though the two regimes interact in practice through training data disclosure and software liability questions.
About Ferraz & Whitmore
Ferraz & Whitmore is an international law firm based in Lisbon, advising business clients across 46 jurisdictions. Our AI and technology law practice supports technology companies, investors. Additionally, in-house legal teams operating in Germany and across the EU on the full spectrum of IP strategy. AI Act compliance, technology licensing. Additionally, digital services regulation questions arising from generative AI deployment. As an international law firm in Germany and throughout Europe, we combine Portuguese civil law expertise with English common law tradition to deliver cross-border legal solutions that work across multiple regulatory systems simultaneously. Our intellectual property team includes practitioners with experience advising on copyright disputes, trade secret protection, and software liability before German courts. The firm's Lisbon base provides direct access to EU regulatory developments, while our common law capabilities support enforcement and arbitration strategies in English-speaking jurisdictions. Ferraz & Whitmore participates in cross-border practice groups focused on AI regulation and intellectual property across civil and common law systems. To discuss how the German IP regime and EU AI Act obligations affect your AI content strategy, contact us at info@ferrazwhitmore.com.
Disclaimer: This publication is provided for informational purposes only and does not constitute legal advice. The information herein should not be relied upon as a substitute for professional legal counsel tailored to your specific circumstances. Ferraz & Whitmore assumes no liability for actions taken or not taken based on the contents of this material. For advice regarding your particular situation, please contact info@ferrazwhitmore.com.