Who Profits When Machines Create?

In 2024, the generative AI market was valued at over $60 billion and growing at a compound annual rate that makes even seasoned technology investors blink. Companies like OpenAI, Stability AI, Midjourney, and Adobe are capturing enormous value from tools that generate text, images, music, and video. Venture capital has poured billions into AI startups. The stock prices of companies positioned to benefit from generative AI have soared. By almost any financial measure, the AI art revolution has been extraordinarily lucrative — but a crucial question remains largely unaddressed: lucrative for whom?

The financial architecture of generative AI art is structured in a way that concentrates value at the platform level while distributing costs across a diffuse base of creators whose work trained the models. The large language models and image generators that power these tools were trained on billions of images, texts, and other creative works — the vast majority scraped from the internet without the explicit consent or compensation of their creators. This is not a minor detail; it is the foundational economic reality of the entire industry. The raw material of AI art is human art, and the humans who created that raw material have received, in most cases, nothing.

The entire generative AI economy is built on the unpaid labor of millions of artists who never consented to their work being used this way. — Concept Art Association, 2024

To understand who profits from AI art, it helps to map the value chain. At the top sit the platform companies — the firms that develop, train, and deploy generative AI models. These companies capture the lion's share of revenue through subscription fees, API access charges, and enterprise licensing. Their valuations reflect the market's belief that they control the critical chokepoint in the creative production pipeline: the model itself. Below the platforms sit a growing ecosystem of applications, plugins, and services built on top of these models, each taking a smaller slice of the value generated.

What is most striking about this value chain is the near-complete absence of the people whose creative work made the technology possible. The artists, photographers, illustrators, and designers whose work constitutes the training data occupy a position analogous to the agricultural laborers in a global food system: essential, but largely invisible and poorly compensated. This is not a sustainable model, economically or ethically. As legal challenges mount and public awareness grows, the current extractive approach to creative labor in AI is likely to evolve — the question is whether it evolves toward fairness or simply toward more sophisticated forms of extraction.

The question of how to compensate artists whose work trained AI models is one of the most complex and contentious issues in the creative economy. Several high-profile lawsuits — including class actions by visual artists against Stability AI and Midjourney, and suits by major publishers against OpenAI — are testing the legal frameworks around training data and copyright. The outcomes of these cases will shape the economic landscape of AI art for decades. But the legal questions, while important, are only part of the picture.

Beyond the legal framework, there is a practical challenge: even if courts determine that artists deserve compensation for training data use, implementing a fair payment system is enormously difficult. How do you attribute value to a single image among billions used for training? How do you compensate an artist whose style influenced a model's outputs without their specific images being directly copied? These are not merely technical questions — they are philosophical ones about the nature of influence, inspiration, and originality that human cultures have grappled with for centuries, now rendered urgent by the scale and speed of AI systems.

Some companies have begun voluntary compensation programs. Adobe's Firefly model was trained exclusively on licensed content, and the company has established a fund to compensate contributors. Shutterstock negotiated a deal with OpenAI to license its image library and shares some revenue with contributing photographers. These early efforts, while imperfect, suggest that a more equitable model is at least theoretically possible. The challenge is extending these principles across an industry that has largely been built on the assumption that online creative content is free raw material.

If the current model of AI art economics is unsustainable, what alternatives might emerge? Several promising approaches are being explored by researchers, advocates, and forward-thinking companies.

• Consent-based training: Models trained only on content for which explicit permission has been granted. Adobe Firefly and some open-source projects already follow this approach, though it limits the diversity of training data.
• Revenue sharing: Platforms distribute a percentage of subscription revenue to artists whose work was used in training, proportional to their contribution. This is technically challenging but not impossible, and mirrors models already used in music streaming.
• Collective licensing: Industry-wide licensing frameworks, similar to those used in music (ASCAP, BMI), that create standardized mechanisms for compensating creators whose work feeds AI systems.
• Opt-in marketplaces: Platforms where artists can choose to license their work for AI training at rates they set themselves, creating a genuine market for creative training data rather than an extractive one.

None of these models is perfect, and each involves significant trade-offs between ease of implementation, fairness to creators, and the advancement of AI capabilities. But they represent a meaningful departure from the current status quo, in which the vast majority of creative value flows upward to platform companies while the creators of the raw material receive little. The most likely outcome is a hybrid approach that combines elements of several models, shaped by legal precedent, market pressure, and evolving social norms around AI and creative labor.

The path forward requires action on multiple fronts — legal, technological, and cultural. On the legal front, clearer frameworks around training data rights, fair use in the context of AI, and artist compensation are urgently needed. The current patchwork of lawsuits and voluntary programs is inadequate for an industry growing this rapidly. Legislators in the EU, the US, and elsewhere are beginning to grapple with these questions, but the pace of regulation lags far behind the pace of technological change.

On the technological front, better tools for tracking provenance, attributing influence, and managing consent are essential. Blockchain-based provenance systems, content credentials standards like C2PA, and watermarking technologies all have roles to play — though none is a complete solution on its own. The goal should be a system in which the creative lineage of AI-generated work is transparent and the contributions of human creators are visible and compensable.

We can build an AI art economy that works for creators, not just platforms. But it requires intentional design, not just market forces. — World Economic Forum, 2024

Ultimately, the question of who profits when machines create is a question about values, not just economics. Do we believe that creative labor has inherent value? Do we believe that the people whose work makes AI possible deserve fair compensation? Do we want an art world that is more equitable or more extractive? The answers to these questions will determine not just the economics of AI art, but its cultural legitimacy. An AI art ecosystem built on fairness and consent will produce better art, attract more talented creators, and earn the public trust that the current model is rapidly eroding.