Who Owns AI-Generated Output? — Insights from Our “Crafting the Future: AI, Creativity & Rights” Workshop

Who owns the works created with AI tools (e.g., Midjourney, Dall-E, LookX)? Are they protected by copyright? To explore this complex issue, we organized a two-day workshop involving young creators and early adopters of Generative AI. We gathered insightful empirical data that sheds light on how the new social contract between humans and AI should look like. In this article, we are sharing key insights and examples from our sessions, including thought-provoking exercises that reshaped our understanding of creativity and ownership in the age of AI.

This post is Co-authored by Negar Kalantar, Ph.D., Ashley Greenwald & me, Paul Jurcys.

Workshop Objectives

In Spring 2024, Negar Kalantar, Ashley Greenwald, and I, Paul Jurcys, had a coffee chat about the impact of AI tools on the creative efforts of artists, designers, and industry professionals (e.g., architects and product designers) and how these tools are likely to shape the future of applied creativity.

While we acknowledged the initial excitement around AI, we recognized that image-generation tools have yet to make a significant impact on professionals in architecture and interior design. One key reason is that these AI tools are not compatible with the software programs and workflows widely used in the AEC industry.

We also encountered another issue — the stance of the U.S. Copyright Office , which currently denies copyright protection for works created with the assistance of AI tools. In three recent cases, the Copyright Office ruled that works produced with AI tools in the creative process are ineligible for copyright protection. According to this position, works co-created with AI are not considered worthy of copyright, regardless of the time and effort invested by creators.

Furthermore, in 2023, the U.S. Copyright Office conducted a study on “Copyright and Artificial Intelligence,” inviting interested parties to respond to an extensive questionnaire on the impact of AI on creativity. The Office received over 10,000 comments; however, one year later, there has been no response from the U.S. Copyright Office addressing the most pressing issues related to the copyrightability of works co-created with AI tools.

Negar, Ashley, and I agreed that more empirical insights are essential to advance this discussion at least in two dimensions:

1. Creativity & AI: How emerging AI tools affect the creative process
2. A New Social Contract: Clarifying the Contours of Ownership and Rights

Therefore, it seemed natural and timely to organize a workshop with architecture and design students, alongside industry leaders, pioneers, and other stakeholders, such as AI tool developers, copyright law experts, and AI governance professionals.

This workshop was an experiment in exploring what it takes to cultivate a true sense of ownership over creative works, recognizing the many gray areas surrounding copyright and ownership.

Day 1 of the Workshop: Exploring the Impact of AI Tools

On the first day of our workshop, we held a closed session with students from Negar’s class at CCA. Negar Kalantar began with a holistic overview of AI tools and their role in reshaping the work of designers and architects. Paul Jurcys then introduced core principles of copyright, including the requirement of human authorship and the idea-expression distinction, emphasizing how these deeply rooted legal concepts are being challenged by emerging AI technologies. Ashley Greenwald brought these ideas into a more concrete frame, providing real-world examples of how issues of creation, ownership, and licensing play out every day in the architecture and design industries.

Through these layered perspectives, we aimed to spark curiosity while anchoring students in the legal and ethical complexities of AI-driven work.

The practical component was equally structured and ambitious. We introduced students to a generative AI tool — LookX — and tasked them with three distinct exercises, aiming to highlight different dimensions of human-AI collaboration:

1. “Prompt-to-Image” — Students were asked to explore the possibilities with only prompts, where they generated an image solely through textual prompts, pushing them to consider the nuances of translating language into visual art.
2. “Sketch-to-Image” — This task required students to start from their own sketch or wire-frame 3d model and use LookX.AI to situate it within a context of their choice, blending manual artistry with AI’s interpretative capability.
3. Creative Collaboration with AI — The final assignment was a free-form collaboration with LookX.AI, giving students full creative autonomy to produce work limited only by their imagination, manual ingenuity and the AI’s range.

These tasks were crafted to help students engage with AI on multiple levels, from directive input to adaptive co-creation. They were given three days to work on these assignments and presented them during day 2 of the workshop.

To deepen the exploration of “ownership” and “copyrightability,” we asked students to assess their experiences with each task closely. We encouraged them to consider how much control they felt over the creative process, the degree of ownership they perceived over the final product, and whether they felt that AI had contributed as a collaborator or merely as a tool.

In particular, we asked the students to document the time spent on each iteration, the number of iterations made, the prompting history (initial and final prompts), and to evaluate their perceived ownership of the final work along with their satisfaction with the collaborative process involving AI.

By providing specific criteria for these assessments, including quantitative insights where possible, we aimed to ground their reflections in data, fostering a more nuanced conversation about authorship, creative autonomy, and the evolving role of AI in artistic and professional practices.

Day 2: Insights on Ownership & Copyright based on Student Presentations

On the second day of the workshop, we opened the doors to the public, featuring insights from prominent industry leaders, including the Keith Krumwiede (Director of the Architecture Department at CCA), LookX’s CTO Charles Liu (thanks for the credits!), Alireza (AL) Borhani, Angela Foss , Omari Barron, Claire Xue, Hilmar Koch, and Chris Gardner. This session provided an opportunity for students to present their works and share their reflections on collaborating with LookX.

Through these presentations and the subsequent discussions, we uncovered a series of nuanced insights into the concept of ownership — a theme that emerged as both complex and deeply layered in the context of AI-assisted creation.

Multiple Dimensions of Ownership

As students were presenting their works, one of the first observations was that ownership has multiple dimensions, each interacting uniquely with AI co-creation:

The most personal aspect, which we call perception of ownership, reflects the creator’s personal connection to their work. Here, the classical copyright concept of a “work as the spiritual child of the author” faces new challenges, as workshop participants questioned the depth of their authorship when an AI played a substantial role in the creative process. This subjective sense of ownership brings forward the notion that the creator’s personal connection to the work may not always align with traditional copyright definitions.

Another layer of ownership emerged as factual entitlement, where ownership is tied to control over the work. In practical terms, this refers to a creator’s ability to decide how the work is used, modified, or shared. In our case, LookX users explored their control over outputs generated with the platform, pondering the extent to which they could direct or alter AI-generated works.

Lastly, we delved into copyright ownership in works co-created with AI. This area raises critical questions, as copyright traditionally requires human authorship — a criterion AI tools challenge. With generative AI like LookX involved, questions about whether these works qualify for copyright protection remain unresolved, adding a legal dimension to the multifaceted nature of ownership in AI-driven creation.

Exercise #1: Prompt-based (text-to-image) Method

The first exercise focused on exploring the nuances of control and ownership over works created using the prompt-based (or text-to-image) method. Eleven talented CCA students used LookX.ai to investigate the possibilities of prompt-driven creativity. In this approach, students crafted visual works by combining their own prompts with the custom functions embedded in LookX.

Students expressed varying views on their level of control over the image creation process and their sense of ownership over the final output. While most students felt they had limited control, with some even stating they had “very little,” others believed they had a greater influence on the creative process and could more readily claim the final work as their own. As one student put it, “The AI is doing what it wants.”

Chris Gardner, an early adopter of generative AI in the visual arts and a leading expert in this field, observed that with increased experience and skill in using AI, new opportunities emerge, and the sense of ownership and control can grow accordingly:

“I create thousands of images a day, and this process is exhilarating. It’s like photography — you sift through ideas, waiting for something to click. And when it does, an architect or a trained eye can see it instantly, relating it to movements like Bauhaus. AI lets me make thousands of variations on that single idea, refining it until the concept is brought halfway to life. And when I say halfway, it’s because AI can’t give us a final concept; it’s the human mind that brings it fully into existence. That, to me, is the essence of ingenuity. All art is derivative; we’re always building on the past. But now, instead of leaning on what others have already done, I’m able to push into new territory. By the end of the day, I’m months ahead of where I could have been on my own.” (Chris Gardner)

Exercise #2: “Sketch-to-Image” Method

The second exercise yielded a markedly different experience concerning control over the creative output. Students who used their own sketches, images, and 3D models as starting points were able to exercise substantial control over the process, including the iterations and final outputs generated with LookX.ai.

In this scenario, all participants agreed that they felt a stronger sense of ownership over the process, believing they should own the generated content more often than not. Almost uniformly, they voiced the opinion that these outputs deserve a level of copyright protection.

Below are three examples of the works created by students:

Exercise #3: Multi-Process Creative Approach

The third exercise for students was to use LookX to “do whatever they want” and collaborate with multiple tools, if they preferred so. Students could have any starting point (a prompt, a sketch, a 3D model), mix and match, photo-edit the works using various digital tools. This assignment was an open opportunity to take a “multi-process approach”.

Here is an example from Ashley’s recent project, where she collaborated with AI tools using a ‘multi-process creative approach.’ Below is a high-level outline:

• Begin with a site walk and capture drone photos.
• Select a set of concept images from various sources.
• Upload the concept images and drone photography, then blend them using Midjourney (resulting in 3,650 variations).
• From these variations, select 10–20 images that are relevant for incorporation into the final concept.
• Download the selected images.
• Upload the drone photography and Midjourney images into GetImg for additional image mixing (yielding 325 images).
• Download the images from GetImg, selecting 5 that are useful.
• Use Adobe Photoshop to combine and cut images from the drone photography, concept imagery, Midjourney, and GetImg.
• Refine the edges of the cut and mixed pieces using the Firefly autofill tool in Adobe Photoshop.

Hours: 30

Outcome: 3 AI-assisted renderings of existing spaces for a client

Time Saved: At least 30 hours (work avoided due to collaboration: no 3D modeling of the space, no floor plans needed, no selection of final materiality, furniture, lighting, or plants)

Originality of Design: Yes, original within the context of the client site

Ownership: Perceived and attributed factual ownership due to the time and effort involved

Thoughts: These tools enable us to think about spaces and products more holistically and reach concepts faster while maintaining originality. They can significantly accelerate the work of seasoned professionals who understand the physics, spatial laws, and the multitude of steps and complexities required to make these ideas a reality. This experience and knowledge guide us in making wise decisions with AI tools, distinguishing between concepts that have structural potential and those that are unrealistic.

A New Social Contract Between Humans and AI

During the course of day 2 of the Workshop, one of the shared perspectives among the participants of the workshop was that creativity is an unfolding process–not a singular act–it is a sequence of ideation, revision, and refinement. It involves countless attempts to polish and perfect, with creators continually iterating interim works until the creator reaches a satisfactory result.

This view of creativity as a layered, evolving process contrasts sharply with the U.S. Copyright Office’s current stance, which appears to simplify the creative process where people use AI tools to a sequence of pressing buttons. Such a view, however, misses the mark: it fails to capture the nuances and complexity inherent in creative endeavors, particularly when AI is involved. Besides, such an approach and risks sidelining the authentic engagement that creators experience when they collaborate with AI tools.

During the workshop presentations and discussions it became clear that as we move forward, it is important to avoid generalizations. One-size-fits-all approach is not only insufficient, it is misaligned with the practices and needs of professionals across creative industries. It also became clear that creative industries would benefit enormously from clearly articulated principles that actually correspond to how creators utilize AI in their creative endeavors.

Looking ahead, we need to think of a “new social contract” — a framework that more accurately reflects the contours of human-AI collaboration. Workshop participants emphasized several criteria that shape their sense of “ownership” and the copyright-worthiness of AI-assisted work. Among these factors are the creator’s ability to influence and control the output, the number of iterations and time invested, and whether the work is generated through prompts or is based on original sketches or images. These are the realities of AI-driven creativity — a set of standards that call for a more sophisticated and inclusive understanding of copyright that can adapt to the demands of modern creators and their tools.

With regard to copyrightability of outputs co-created with AI tools, there is one particular scenario where the copyrightability should not be questioned: this scenario occurs in situations where humans utilize several tools to achieve a final output with which they are satisfied.

The best illustration is Ashley’s collaborative approach described in the previous section. Other professional creators who participated in the workshop — Claire, Chris, Negar — collaborate with multiple tools and create many interactions before the desirable outcome is achieved.

The Emergence of “AI-Enabled Creator”

Negar made an insightful observation on how the relationship between architects & designers and the tools they use has been evolving over time. In the early days of digital transformation, software tools like CAD were seen purely as instruments or tools — valuable to conduct specific tasks. Architects and designers used them to enhance precision and productivity without fundamentally changing their roles.

In the next phase, these tools began to take on a more collaborative role, helping to manage complex processes and enabling deeper, more integrated workflows (one of the most notable examples is BIM).

Now, we are entering a new phase, where AI becomes an embedded partner in the creative process, dissolving the boundaries between tool and creator. AI tools are increasingly becoming not just as aids but as collaborators, almost like extensions of the individual’s creative mind. The line between human and machine blurs as AI weaves itself seamlessly into the process, and the relationship between humans and digital, AI-powered tools is more complex.

Paths Forward

As we look to the future, three key areas of focus emerge.

First, we expected the age of participants to influence how quickly they adopted AI tools. Yet, one striking observation during the workshop transcended age: all creators — whether students, apprentices, or seasoned professionals with decades of experience — found themselves on an unexpectedly level playing field in terms of generative AI skills.

“Humanity has never seen a moment where anyone — from novices to experts with 50 years in the field — can approach a technology from virtually the same starting line.” — Keith Krumwiede

This unprecedented shift offers a clear choice: learn and adapt to new models of collaboration with AI, or risk being left behind.

Second, we see an ongoing need to engage deeply with this evolving domain, observing how AI tools continue to develop and how industries rooted in applied creativity adapt to these changes. We are committed to further workshops, events, and research to push forward the conversation on AI’s role in reshaping creative processes and professional practices.

Contact us if you’re interested in getting involved or partner with us in exploring this rapidly changing domain.

Third, generative AI is beginning to unlock new avenues for job creation and income, especially for creatives. Whether it’s by licensing creative data to companies or exploring new monetization models, the potential exists for more robust compensation within a reimagined creative economy. However, without a clearer legal framework, these opportunities remain stunted.

A more balanced approach from copyright authorities could foster a positive climate for innovation, empowering creatives to fully embrace this new frontier. We will continue leading the conversation around the contours of this “new social contract” between humans and AI.

🙏Thank you for reading!

🌎 Curious to follow our work on Creativity in architecture and design? Subscribe to this newsletter!

🔎 If you’d like to discuss more about AI, Copyright and Creativity, feel free to DM me — or even have a chat with Paul AI, my digital knowledge twin.