Generative AI and Fair Compensation for Artists
Tech companies are beginning to monetize generative AI, prompting creators to demand fair compensation. However, there is disagreement over how much artists should be paid.
The Call for Proper Authorization and Compensation
A recent open letter from the Authors Guild, signed by more than 8,500 authors including Margaret Atwood, Dan Brown, and Jodi Picoult, urges generative AI companies to stop using their works without proper permission or payment. Many artists have filed lawsuits against generative AI vendors like Stability AI, MidJourney, and Microsoft regarding copyright infringement and misuse.
Vendors’ Efforts to Pay Creators
Some vendors have committed to establishing “creators’ funds” to compensate the artists, authors, and musicians whose works were used in developing their generative AI models. Some of these funds have already been launched, signaling a move towards fairer and more sustainable business models for generative AI.
How Much can Creators Expect to Make?
This seemingly simple question becomes difficult to answer when examining the various compensation policies proposed by generative AI vendors. Despite our efforts, determining realistic earnings for creators has proven challenging.
The Challenge of Vague Terms
Generative AI models learn to create images, music, text, and more by analyzing patterns in a large number of examples sourced from the web. Many of these examples are copyrighted or published under specific usage licenses that vendors often disregard without informing the original creators.
Differing Perspectives on Copyright Law
While some companies argue that they are justified in training their generative AI on copyrighted works under the “fair use” doctrine (in the United States), these legal questions remain unsettled. Furthermore, public opinion largely supports creators who often earn significantly less compared to the profits made by tech and AI companies.
Efforts of Generative AI Vendors
Adobe, Getty Images, Stability AI, and YouTube have introduced or promised methods for creators to share in the profits generated by generative AI. However, these companies have been vague about the exact earnings creators can expect. This lack of clarity further complicates the decision-making process for artists considering allowing vendors to train models using their works.
The Compensation Policies of Adobe and Getty Images
Adobe trains its generative AI models called Firefly using images from its stock asset library, Adobe Stock. They offer a once-a-year “bonus” payment based on criteria like approved images used for Firefly training and the number of licenses those images generated. The specific value attributed to each image and license is unclear. Additionally, contributors must reach a $25 minimum threshold before withdrawing their earnings.
Getty Images plans to compensate contributors to its generative AI tool on an annual recurring basis. Contributors will receive a share based on both the assets they’ve contributed to the model training data set and traditional licensing revenue. However, the actual payment structure remains undisclosed.
The Ambiguity Surrounding Shutterstock’s Compensation
Shutterstock provides one-off payments through its Contributors Fund, which distributes payouts twice a year. Payments are proportional to a creator’s contributions to Shutterstock’s content library. Additional compensation may be given if new content produced by Shutterstock’s AI generators includes the creator’s work. Though concrete figures are unavailable, a survey suggests that artists receive an average of $0.0078 per image from the Contributors Fund.
Remaining Uncertainty for Creators
Unfortunately, generative AI compensation schemes remain mostly theoretical. For example, Stability AI’s revenue sharing plan with musicians through AudioSparx is still being developed. YouTube is also in early stages and plans to collaborate with partners in the music industry to create fair monetization models for generative AI.
Lack of Concrete Figures and Transparency
Many generative AI vendors decline to provide exact dollar amounts that creators can expect to earn. Some attribute this absence of data to the novelty of the technology and business model, while others claim the wide range of potential earnings makes it challenging to offer a useful estimate.
Promising Light: Braia’s Creator-Focused Approach
Braia, which exclusively trains its art-generating AI on licensed images, implements a revenue-sharing model based on the impact of contributors’ data. Artists have control over setting prices for each AI training run, providing a more transparent approach compared to other vendors.
The Current Reality for Creators
As it stands, few generative AI vendors present a compelling case for artists to opt into their training programs. Vague promises of future riches fail to address the immediate financial needs of many creators heavily reliant on contractual income.
Disclaimer: The information provided is subject to change as companies continue developing their generative AI strategies and compensation policies.