Image-Generating AI: Trends and Legal Challenges | Holland & Knight LLP
[co-author: Rodrigo Javier Velasco]
Like human intelligence, artificial intelligence (AI) can recognize patterns, predict outcomes, analyze complex problems, and even match – or surpass – human creativity. Technology-generated innovative and creative processes are now a reality, and they come in various forms.
On the visual art and/or photography side, Prisma Labs recently released an intriguing and amusing “magic avatar” app, Lensa AI, that creates artistic and fantastical portraits.
The user uploads a minimum of 10 selfies from which Lensa AI creates a series of computer-generated portraits of the user, incorporating many different styles and themes applied to the user’s facial elements as well as the image’s background (e.g. anime , vintage and many more) ).
According to Lensa AI, these magical avatars are created using an external technology known as Stable Diffusion (stability.ai). In general, this external technology is a deeply structured, machine learning method with the ability to process and generate highly elaborate images. As with any new technology – especially one that manipulates existing images and creates new ones – a variety of legal issues come into play.
A common law approach
With new technologies, developers and users must consider issues of ownership and control of various intellectual property (IP) assets. The creation of AI-generated images and text (such as ChatGPT) can lead to legal issues involving patent and copyright law, protection of one’s rights of publicity and privacy, and ethical issues that will need to be addressed by courts and legislatures.
Without providing an in-depth doctrinal legal analysis for these legal fields, some general takeaways and/or legal questions to keep in mind are described below.
As with any other new technology, it will be important to determine whether images generated by deep learning image generator platforms can be used for marketing and branding purposes.
Even if individual users have not used their own faces for branding purposes or allowed others to do so, a user may eventually see their face on the developer’s website or marketing material by assigning rights to images via an AI application to a developer to allow Potential users should be aware that they may waive their rights of publicity and privacy.
The US Court of Appeals for the Federal Circuit has ruled that AI cannot be listed as an inventor on a patent application. The topic is further addressed in the Holland & Knight blog post, “Revisiting AI Inventorship in Thaler against Vidal.”
Treatment of visual works created through the involvement of AI is under consideration at the US Copyright Office. “Zarya of the Dawn,” a graphic novel by Kristina Kashtanova, was created with the help of the AI image-generating technology MidJourney. On September 15, 2022, the Copyright Office granted a valid copyright registration for such work, and it appears to be the first of its kind (Reg. #VAu001480196). A month later, the Copyright Office notified that cancellation of the registration was possible and asked the claimant to provide details of the creative process to demonstrate that there was substantial human involvement in the creative process.
The Copyright Office will accept works that are partially and not entirely assisted in their creation by AI. These criteria agree with the recent decision Thaler against Vidal.
The patent applicant, dr. Stephen Thaler, had already been unsuccessful in submissions to the copyright office, which refused to register the work, “A Recent Entrance to Paradise,” because no human author was named on the application.
Unlike “Zarya of the Dawn”, which was created with the help of AI, “A Recent Entrance to Paradise” was created entirely by AI. In all likelihood, the line between fuzzy and subjective will be blurred.
Product Liability, Technology Commercial Contracts and the Insurance Industry
Unprecedented irreparable damage to users in various forms may be a risk for the AI companies, including the violation of IP rights and goodwill. The risk allocation behind the use of AI image-generating processes will be key to determining whether the AI companies are potentially liable. New types of AI-related insurance for the companies/individuals behind the image generation or any AI ecosystem can be developed to address the IP issues as well as cases of massive data breaches and bad faith uses.
For more information
AI methods, including technology-generated processes that create images, are likely to have profound and unforeseen impacts on our everyday activities. For this reason, legal and even ethical questions must be considered by developers and users.