Gerald Carter built Diversity Photos from scratch starting in 2016, curating 12,000 stock images focused on representation in minority communities. When Adobe began feeding those images into its AI models, including Firefly and Adobe Sensei, Carter asked for one of two things: either exclude his library from training data, or negotiate a proper licensing deal with fair compensation.
Adobe offered him $1,173.93 as a "bonus fee." Carter, whose licensing rates for AI evaluation run into the hundreds of thousands for a single year, rejected what he called an insulting offer.
So he took it to arbitration. He lost.
The Contract Loophole
The core of Adobe's defense was a Stock Contributor Agreement Carter signed in 2018. That contract permitted Adobe to use uploaded images for "developing new features and services [to promote my work]." Adobe argued the word "new" was broad enough to cover AI training, even though generative AI products like Firefly didn't exist when Carter signed.
The arbitrator agreed, granting Adobe's summary judgment motion on most of Carter's claims. His strongest remaining argument, a negligence claim that Adobe failed to watermark his images (which allowed other AI companies like Midjourney, Stability AI, and Google to scrape and train on them without consent), also failed.
Carter eventually withdrew his last remaining claim. The reason was bluntly practical: he couldn't afford the $24,000 in arbitration costs to continue.
Pre-AI Contracts as Blank Checks
The timeline tells the real story here. Carter signed his agreement in 2018. Adobe launched Firefly in March 2023. Carter sent a formal demand in January 2024. He filed for arbitration in June 2024. A process server showed up at his door in November 2025.
Adobe is now seeking court confirmation of the arbitration award, which would turn it into binding legal precedent. That's the part that should concern every creative professional who uploaded work to any platform before 2022.
The practical takeaway: contracts written years before generative AI existed are being interpreted to cover AI training. The language doesn't need to mention "artificial intelligence" or "machine learning" explicitly. Broad terms like "new features and services" are enough, at least according to this arbitrator.
For Adobe Stock contributors specifically, this ruling suggests the company believes its existing agreements already grant permission to use contributor images for AI training. Contributors who assumed they could opt out may find that their contract says otherwise.
Carter's case also highlights a structural problem in how these disputes get resolved. Mandatory arbitration clauses push creators into a process that costs tens of thousands of dollars, far out of reach for most individual photographers and artists. Adobe, with its $200+ billion market cap, faces no such constraint.