Exploring the Impact of Adobe Firefly on Artist Copyright Issues
Written on
Chapter 1: Introduction to Adobe Firefly
Adobe Firefly, initially launched with the Photoshop (Beta) app, is set to roll out to the public version of Photoshop later in 2023. Although it is not yet available for commercial purposes, the tool currently supports text prompts in English, with plans for broader inclusivity in future updates.
Firefly is Adobe's innovative text-to-image model, which has made a significant impression, resulting in nearly 200 million AI-generated images since its debut in March. Due to its popularity, Adobe plans to extend Firefly's capabilities to its 12,000 enterprise clients.
Section 1.1: Data and Usage Concerns
The training data for Firefly includes over 100 million images, comprising Adobe's stock, licensed, and public domain content. Adobe maintains a vast collection of quality stock images sourced from contributors, who typically earn 33% of royalties from sales or usage. Recently, Adobe Stock introduced a new category for AI-generated images, but only accepts submissions from contributors with the proper rights.
Concerns have been voiced by numerous Adobe Stock creators about the utilization of AI-generated images for training Firefly. Many creators feel they lack the option to exclude their work from being used in this capacity. Adobe's licensing agreements allow for such training, yet the company has assured creators of future compensation once Firefly exits its beta stage, though the details remain unclear.
Subsection 1.1.1: The Copyright Dilemma
As creators increasingly turn to generative AI tools, like text-to-image models, they have become more vigilant about copyright issues regarding training content. To mitigate these concerns, Adobe aims to provide assurances of legal safety and protection. Measures have been instituted to block specific copyrighted, violent, and sexual keywords in text prompts, and the company's terms of use strictly forbid the generation of abusive, illegal, or confidential material.
Chapter 2: The Generative Fill Tool
Adobe recently introduced its Generative Fill feature, powered by Firefly technology. Marketed as "the world's first co-pilot in creative and design workflows," this AI tool enhances Photoshop's functionalities, integrating with existing features like neural filters and Content-Aware Fill. Users can now leverage text prompts akin to those in Dall-E and Midjourney to add, expand, or eliminate unwanted elements from images.
Adobe Firefly (Beta) - Generative Image AI review - YouTube
This video reviews Adobe Firefly, highlighting its features and potential impact on creative workflows.
As Adobe has been incorporating AI into its tools for over a decade, the inclusion of Firefly in Photoshop significantly broadens the software's capabilities. Users can now create images and templates from scratch, making photo editing faster and more efficient. The tool allows users to generate objects, develop new backgrounds, extend images, and remove unnecessary elements, all with simple text prompts.
However, this advancement raises critical questions about copyright concerning the newly expanded images. Does the copyright protection of the original image extend to new images created using Generative Fill? Essentially, if someone employs Generative Fill on an existing photo to create a larger composition, can they claim it as a unique artwork?
Section 2.1: Legal Perspectives
Thomas Maddrey, Chief Legal Officer and Head of National Content and Education First at the American Society of Media Photographers (ASMP), addresses the copyright complexities inherent in this technology. He highlights concerns about ownership and copyright protections concerning images used for training AI models. The question remains whether outputs generated by Generative Fill can be classified as new works under the current copyright law when combined with pre-existing photographs.
ASMP asserts that photographers have valid grounds to challenge unauthorized use of their work to create new AI-driven content. However, the fair use defense may complicate matters, as illustrated by the Warhol v. Goldsmith case, which raised important issues regarding the qualifications of AI-generated expansions as fair use. The definitive application of this ruling to generative AI remains ambiguous until a legal case is brought forth.
Print on Demand Design AI Showdown: Adobe Firefly vs Midjourney V5
This video compares Adobe Firefly with Midjourney V5, testing their capabilities in the realm of generative design.
The US Copyright Office has refrained from commenting on specific AI technologies but refers to its policy statement regarding copyright registration for works incorporating AI-generated elements. Current guidelines mandate that any portions of an image created through generative AI must be disclosed during the copyright registration process.
Moving Forward
It is clear that rules governing generative AI art will continue to evolve as new copyright challenges emerge. This ongoing debate represents a significant struggle between artists and some of the largest technology companies globally.
Ginger Liu, the founder of Ginger Media & Entertainment, is a Ph.D. researcher focusing on artificial intelligence and visual arts media, as well as an author, journalist, artist, and filmmaker. Tune in to the podcast for more insights.