AI-Generated Representations of Architectural Works and Limits of Architectural Copyright

Why is architectural copyright quite limited?

Architectural works are different from other copyright works: they serve functional purposes and exist in public space. The extent of copyright protection over works incorporating technical solutions should be low, given that those features lack originality. In addition, the public should be able to engage with architecture as part of public space in ways that do not infringe upon authors’ moral and economic rights. Interestingly, despite the Berne Convention recognizing copyright for architectural works in 1908, the United States resisted extending such protection to architecture for a long time. The U.S. only adopted the Architectural Works Copyright Protection Act (AWCPA) in 1990. It aimed at maintaining a pragmatic approach suited to the nature of architectural works.

Protection for architectural works is now codified at 17 U.S.C. §102(a)(8). As defined in 17 U.S. Code § 101, an “architectural work” is “the design of a building as embodied in any tangible medium of expression, including a building, architectural plans, or drawings”. This definition includes the building’s overall form and arrangement as well as composition of spaces and elements in the design but specifically excludes “individual standard features”. Individual standard features are considered to include windows, doors, and other common building components, as further explained in Circular 41 of the U.S. Copyright Office on Copyright Registration of Architectural Works. 

Addressing the second intrinsic characteristic of architectural works — their public nature — Section 120(a) provides that copyright in a constructed building “does not include the right to prevent the making, distributing, or public display of pictures, paintings, photographs, or other pictorial representations of the work, if the building in which the work is embodied is located in or ordinarily visible from a public place.” This creates a statutory “freedom of panorama” for architecture in the United States. The exception addresses the special relationship between architecture and society, which naturally views architectural works as an intrinsic part of shared culture. Society thus gains the right to interact with these works through various forms of expression without infringing the architect’s copyright in the building itself. 

Why are AI-generated copies of architectural works likely to be deemed non-infringing? 

So how does the freedom of panorama apply to AI-generated images of architecture? To answer this, we need to identify which exclusive rights the panorama exception covers and the subject matter to which it applies. As provided by Section 120(a), copyright protection afforded to architectural works hence does not limit the making, distribution or public display of pictures, paintings, photographs, or other pictorial representations of such work, provided the work is located in or ordinarily visible from a public place. Importantly, by including “other pictorial representations of the work” in the panorama exception, without specifying the means of making them, the law remains technology-neutral.

Do AI-generated images qualify as pictorial representations? Clearly, yes. Did lawmakers anticipate GenAI when drafting this provision? Almost certainly not. But here is the key point: given how reluctantly and belatedly the US extended copyright protection to architecture in the first place — a stance that reflects a broader philosophical preference for economic considerations over morality-based justifications in copyright law — the US copyright system will likely take an equally permissive stance toward AI-generated pictorial reproductions. In practical terms, this means AI-generated images of existing architecture fall squarely within the panorama exception. From a policy perspective, it is submitted that the concept of freedom of panorama, when applied to AI-generated pictorial representations of architectural works, continues to help achieve the intended level of copyright protection in the US.

How permissive is freedom of panorama across EU jurisdictions?

In the EU, Member States may implement the freedom of panorama exception under Article 5(3)(h) of the InfoSoc Directive, which permits the reproduction and communication to the public of works of architecture or sculpture that are permanently situated in public places. However, the picture across Europe is fragmented. The Directive allows but does not require Member States to enact such exceptions, and as a result, the rules vary considerably from country to country.

For example, Germany has implemented the freedom of panorama exception in Section 59 of the Act on Copyright and Related Rights, allowing anyone for any purpose to reproduce, distribute and make available to the public works located permanently on public roads or open spaces. France, by contrast, introduced a more restrictive exception in Article L122-5, 11° of the Intellectual Property Code, limiting such reproductions to private individuals acting for non-commercial purposes. In Greece, under Article 26 of Law 2121/1993 on Copyright, Related Rights and Cultural Matters, while not restricted to non-commercial uses, communication of images of architectural works is allowed by mass media only. In fact, many freedom of panorama provisions across the EU draw a distinction between commercial and non-commercial uses, but rarely restrict them to certain actors only. In any case, all these differences add further layers of complexity in the context of generative AI.

Therefore, while the approach adopted by the InfoSoc Directive does seem rather permissive, national approaches to freedom of panorama vary significantly, which makes it difficult to talk about a unified EU approach in relation to AI-generated reproductions of architectural works at this stage.

*Mariam Aroian is a guest researcher at the Institute for Globalisation and International Regulation (IGIR). She is an intellectual property law researcher specializing in the intersection of modern technologies and legal governance. She holds an LL.M. in Intellectual Property and Technology Law from the National University of Singapore.Her research focuses on global IP governance and ethical frameworks for emerging technologies. She has published and presented on intellectual property, cultural rights, and art law, with a particular focus on how these concepts evolve in response to modern technologies and the development of AI.