AI generators borrow from thousands of years of human creativity to produce new artwork. But this can also be an issue: who owns the rights to these artworks?
Legislation is still playing catch-up, but some regions have already provided some guidance. According to the US copyright office, you can’t copyright AI images as they are not original works of authorship by a human being. But that doesn’t mean you can just use them however you want. There may be other legal restrictions in place such as trademarks, patents, and existing platform contracts with its users.
So, before you go and sell your AI masterpiece for millions of dollars, make sure you check the terms and conditions of the AI platform you used, consult a lawyer if necessary, and understand the legal implications involved.
This article attempts to discuss various issues as they pertain to AI image rights. We’ll explain why such issues exist, potential ethical concerns, and more hot-button issues, Let’s get started:
What Are AI Images and How They Are Created
AI images are pictures that are created by algorithms that can combine text and visual information. They are often used to generate realistic or artistic images from a description in natural language.
It’s not the prettiest picture in the world, but straightaway you can tell that AI images are not always original. Most of the time, the algorithms are fed with existing artwork to know how to draw certain styles or concepts. It’s often without the owner’s permission or recognition, leading to questions such as who owns the rights to the images and whether they infringe on the original artists’ copyrights. So…
Who Owns the Copyright in AI-Generated Art?
Under ordinary copyright law, the creator or artist assumes the ownership and rights that determine how the images are used. With AI, it’s a little tricky as there are multiple contributors to the creation of the image. For instance, the rights may as well be granted to the algorithm developer, who designed and programmed the software that produces the artwork. Or, it could be the cumulative data providers that supplied the images, texts, sounds, or other inputs that the algorithm uses to generate the output.
Users also play a part by selecting the parameters, settings, and prompts that influence the final result. The larger public might as well be the rightful owners as no human authorship or originality can be attributed to AI-generated artwork in certain regions such as the US.
As of now, this remains an open question and is subject to commentary and regulation by different copyright offices around the world.
At least with stock images, it’s clear that the original creator has the rights, and they transfer them to others by uploading the images to free stock image sites.
AI Rights As Seen by Different AI Image Tools & Governance Institutions
The US Copyright Office
The position of the USA Office of the Copyright on AI image rights is that only human authors can claim copyright protection for their original works of authorship. It means that AI-generated images, such as those created by neural networks, are not eligible for copyright registration unless they incorporate some human input or intervention. The Office also states that it will not register works produced by a machine or a mere mechanical process that operates randomly without any creative input and intervention from a human author.
Does a human prompt suffice as human artistry? Well, according to the latest guidance from the U.S. Copyright Office, an image generated solely from a text prompt does not qualify for human authorship. The image cannot be registered for copyright protection, as it lacks the traditional elements of authorship a human user would contribute.
The text prompt is analogous to instructions you may provide to a commissioned artist, such as a magazine editor hiring a photographer or the Arte della Lana commissioning Michelangelo to carve the twelve Apostles. During the creation of the art itself, the human user doesn’t ultimately determine how the AI interprets or understands the prompt and generates the final art. It, therefore, follows that the AI and the commissioner are not co-authors in the work.
That said, if you happen to significantly modify the AI-generated image in a sufficiently creative way, for instance, using Adobe Photoshop or other tools to introduce new hand-drawn elements, then the modifications may be eligible for copyright protection. Similarly, you can be deemed to have contributed sufficient human artistry and creativity by arranging the AI-generated images in a specific sequence. For instance, you can use art to design the layout for a comic book, which may subsequently be eligible for copyright protection.
Midjourney AI
Midjourney is a service that allows you to create and edit digital assets using various tools and features. However, the ownership and usage rights of these assets depend on whether you are a paid member or not. If you are a paid member, you have full ownership of the assets you create with the service, except for the ones that are upscaled from other sources. In that case, the original creators retain their rights.
If you are not a paid member, you do not own the assets you create with the service. Instead, Midjourney grants you a license to use them for non-commercial purposes only, and you must give credit to Midjourney as the source. This license is based on the Creative Commons Noncommercial 4.0 Attribution International License.
Australia
Australia has a clear policy on AI-generated work: it does not recognize it as eligible for intellectual property protection. According to the Australian Patent Office, AI-generated inventions are not patentable because they do not meet the requirement of having a human inventor. Similarly, AI-generated works are not protected by copyright under Australian law because there is no registration system for such works and they do not fall under the existing categories of authorship.
Open AI – DALL.E
DALL·E is a powerful tool that can generate images from text. You have ownership of the images you create with DALL·E, whether you use free or paid credits. You can use them for any purpose, such as printing, selling, or merchandising.
You’re encouraged to respect the Open AI’s Content Policy and Terms when using DALL·E. That is, you should be honest about the AI involvement in your images. For instance, while you can remove the DALL·E logo in the bottom right corner if you want, you cannot pretend that the images are human-made or real photographs.
When Can AI Images Be Copyrighted?
The general rule is that there must be some significant human input in the creation of the image, not just providing prompt or setting parameters. Otherwise, the image may be considered derivative work that is not eligible for protection and could actually be infringing on the existing artwork. Additionally, simply transforming an image with AI does not make it original or creative enough to qualify for protection.
To make AI images copyrightable, one possible solution is to use AI as a tool to generate images and then use human creativity to modify and combine them into new works. For example, one can extract features from an AI-generated image and add them to another piece. This way, the human input is more substantial and the final work is more likely to be original and expressive.
For instance, if you provide a prompt to create the following image. The AI output is not copyrightable.
It can’t be copyrighted.
But if you extract some elements from the AI image and hand-paint them into another image, you can make the case that the new artwork can be copyrighted.
The final image with hand-painted elements is copyrightable
In another scenario, you can first draw your own prompt, and use it as a prompt in the AI image generator. You can make a good case that the final image should be copyrightable.
A hand-drawn image transformed using AI to a final complete image that may be copyrightable.
There Is a Difference Between Copyrighted Work and Copyrighted Characters
One of the common misconceptions about intellectual property rights is that they only apply to the work itself and not to the characters depicted in it. However, this is not true. Characters that have been created and copyrighted by someone else are also protected by law and cannot be copied or used without permission.
So, if you draw a new picture of a character doing something different, you are still infringing on the original creator’s rights. You cannot claim that your art is original or fair use if it is based on someone else’s character.
What About Characters of Images in the Public Domain Images
Using AI to create new works based on public domain characters is a creative and innovative way to explore different possibilities and interpretations. Public domain characters are characters that are open and available for use by the public without worrying about copyright infringement. Some famous public domain characters include Sherlock Holmes, Dracula, and Robin Hood.
Still, you can’t claim exclusive rights over the public domain character or the original work that they are featured in. The public domain character remains free for anyone else to use and adapt in the way they see fit. Therefore, you should always acknowledge the source of the public domain character and respect the original creator’s vision and intention.
Additionally, you should use your human artistry to enhance and refine the AI-generated work, as AI may not always produce accurate or coherent results. You should also ensure that your AI-generated work does not infringe on any trademarks or other intellectual property rights that may be associated with the public domain character. For example, you can use Mickey Mouse as he appeared in Steamboat Willie (1928), but not as he appeared in Fantasia (1940) or any later versions.
Potential Ethical Issues and Challenges Around the Use of AI-Generated Work
AI-generated images pose some ethical challenges for human creators and consumers. Some of the potential ethical issues and difficulties pertaining to AI-generated art include:
- AI art can be produced quickly and abundantly, which may reduce its value and uniqueness;
- As image generators continue being more refined, human art may fail to compete with or complement AI art in terms of originality and quality;
- It will become increasingly difficult to verify the authenticity and quality of human-generated work;
- It will be necessary but difficult to ensure that human artists are not plagiarizing or copying AI-generated work, or vice versa;
- There will be a need to protect the intellectual property rights and moral rights of both human and AI creators;
- Human artists will need to disclose the use of AI in the creative process;
- It will become important to establish a clear system to distinguish between human-only, human-AI collaboration, and AI-only work.
- Consumers will need to be informed about the level of AI involvement and influence in the work;
- There will be a need to respect the artistic intentions and expressions of human creators who wish for their art not to be scrapped or used to train algorithms.
The Way Forward
AI image rights will continue to be a hot topic, as more and more AI tools create images for humans, drawing inspiration from human art. Some of the possible solutions that will attempt to resolve the ethical and legal issues surrounding AI-generated art will include:
- Developing clear and transparent guidelines for the use and attribution of AI images, such as Creative Commons licenses or digital watermarks.
- Creating ethical standards and codes of conduct for AI artists and researchers, such as respecting human dignity, diversity, and privacy.
- Establishing legal frameworks and regulations for AI image rights, such as defining the authorship, ownership, and liability of AI-generated art.
- Encouraging collaboration and dialogue among stakeholders, including human artists, AI developers, policymakers, and consumers.
If you’re just into AI to create free images for your blog, you can look into getting stock images from popular sites such as Unsplash, where the licensing is more straightforward.
Frequently Asked Questions
What Are the Existing or Proposed Mechanisms for Licensing, Watermarking, or Tracing AI Images?
There are different existing or proposed mechanisms for licensing, watermarking, or tracing AI images, such as
- Using digital asset tracking solutions that can monitor the online usage of images and detect unauthorized or non-compliant copies.
- Embedding invisible watermarks or signatures into AI images that can identify the source and ownership of the content.
- Developing legal frameworks and agreements that define the terms and conditions of using AI images for commercial or non-commercial purposes by AI platforms.
- Relaxing existing guidelines on the use of AI image art and elaborating on the rights bestowed on AI platforms, as their place in the modern world is imminent.
Can AI Images Be Distinguished From Real or Authentic Images?
Artificial intelligence (AI) can generate realistic images that are hard to distinguish from real ones. However, there are some methods that can help identify AI images, such as looking for inconsistencies, artifacts, or patterns in the image. For example, AI images may have blurry or distorted edges, unnatural colors or lighting, or repeated textures or shapes. They may be signs that the image was created by an algorithm rather than a camera or human author.
Can I Copyright AI Work?
Yes, there have been copyrights and patents issued for different works created by the use of AI. It’s just that copyrighting AI work is a complex and evolving legal issue. Now, here are some general steps to follow when seeking to protect your AI work:
- Identify the type of work you have created or used with AI, such as software, data, algorithm, model, image, or output..
- Determine the ownership and authorship of the work, such as whether it is original, human-made, collaborative, or derivative.
- Assess the level of protection you need or want for your work, such as commercial, academic, or public domain.
- Choose the appropriate form of intellectual property (IP) protection. For instance, you can get a patent, trademark, or copyright.
- Register your work with the relevant IP authorities or agencies in your country or region, if necessary.
- Enforce your rights and respect the rights of others when using or sharing your work with AI.
What Are the Best AI Image Generators?
Some of the best AI image generators include DALL E – 3, Dream Studio, Stable Diffusion Web, Night Cafe, Microsoft Designer, and Imagen 2 in Gemini. We have a more comprehensive review of the tools, noting how they have performed.
