Most of us don’t think about copyright often in our daily lives. But in the era of generating AI, it has quickly become one of the most important issues in the development and output of chatbots and image and video generators. This is something that affects all of us because we are copyright owners and authors.
Sadly, copyright and artificial intelligence are a mess. The race to develop state-of-the-art AI models shows no signs that it will slow down any time soon. To create those next-generation models, tech companies are looking for a lot of high-quality human-generated content. They need these works to make their AI models better, whether it is the personality that makes the chatbot more lifelike or the art style that the image generator is more referenced. AI enthusiasts, on the other hand, may wonder if copyright protection can be obtained for creative works supported by AI.
Most AI companies are vague about what to use, which has led to more than 30 lawsuits that have left them struggling in U.S. courts. You may have heard of some of the most notable people, such as The New York Times v. Openai, where publishers accused Chatgpt of using reporters’ stories verbatim without proper attribution or permission. (Disclosure: CNET's parent company Ziff Davis filed a lawsuit against OpenAI in April, accusing it of infringing on Ziff Davis' copyright in training and operating its AI systems.) Meta has also recently been stuck in hot water, as reported by the Atlantic, and has released a search database of all copyrighted and potential books, which the company does not need to train the company's training.
I spent a lot of time thinking about copyright and AI in my work reporting on AI creative services. I have interviewed intellectual property lawyers, spoke with many creators involved, and spent too much time breaking down the law from government agencies. I've used this experience to make a guide about what you need to know about the era of copyright era, and we'll keep it updated as things change.
What is copyright?
Under the Copyright Act of 1976, copyright is a set of rights to expression that protects “original works fixed in any tangible medium of expression, now known or later developed, from which they can be perceived, reproduced, or otherwise conveyed”.
In other words, copyright is a legal protection that gives the original author rights and the right to control his original work. Copyright protection can be applied to books, art, music, movies, computer programs, blogs, architectural design, drama, choreography, and more. We are all copyright owners. As the U.S. Copyright Office says: “Once you create the original work and fix it, it’s like taking a photo, writing a poem or blogging or recording a new song, you’re the author and the owner.”
There are two ways to intersect copyright with AI. In terms of output, people using artificial intelligence services such as chatbots and image generators want to know if their AI-enabled work is eligible for copyright protection. There are many concerns about the illegal use of copyrighted materials in terms of development. That's all we know so far.
Can I use AI-generated images or text with copyright?
As with many legal questions, the answer is: it depends.
Our guidance on this issue mainly comes from what the United States can do, and we are responsible for managing copyright. The office has released a series of reports on AI and copyright under its latest guidance. In the second report, the office maintained its position that images and videos generated by AI are not in compliance with copyright protection.
However, there are many generator AI editing tools available now. These tools are not created wholesale, but they use AI Gen Gen to add or delete objects, De-Age actors, or perfect audio and video. You can still register and have the potential to get copyright protection for AI-edited content, but you must disclose your use of AI. In the Public Records portal, you can see in the comments how people use AI in creating copyrighted work.
Can copyrighted content be used to train AI?
The basic premise of copyright law is that rights holders – usually the original creator and sometimes in other cases can be one’s employer – can decide how they wish to use their work. In many cases, the owner chooses to license its content; this allows people to charge work appropriately attributed to copyrighted protection. So if the copyright owner wants to allow AI companies to use their content to train AI models, there is nothing wrong or illegal about it. Many publishers, including the Financial Times and the Axel Springer brand, have struck multi-million dollar deals with AI companies to do that.
Problems arise when an AI company potentially uses copyrighted content when it is licensed by the copyright holder. That's what happened to the creators in many lawsuits, including a class action lawsuit against stability AI by concept artist Karla Ortiz. Currently, there are more than 30 proactive lawsuits between AI companies and creators on copyright issues.
Decades of copyright law precedent say that such use is not allowed without permission. Some creators claim that tech companies infringe on their copyrights. Infringement occurs when the permission of the copyright holder defined by the Office of Copyrights is “copying, distributing, publicly displaying, or making into derivative works without permission” “without the permission of the copyright holder.”
The court will decide whether the use of copyrighted materials in AI development meets the threshold for infringement. Meanwhile, many tech companies are trying to find another solution: the fair use exception.
What is fair use and what does it have to do with AI?
The theory of rational use is a basic part of copyright law, which is part of the copyright law of 1976. Fair use allows people to use copyrighted content without the express permission of the holder. In the era before AI, fair use cases included a teacher, using a copyrighted book, or journalists citing copyrighted work in news reports. There are four factors that help determine whether someone’s use can be used for a reasonable purpose, including:
-
Purpose of use: How will people using copyrighted materials use it? Business interests – whether you can make money from use – are important here.
-
The nature of copyrighted work: What is the actual format for a controversial job? Is it a fact like a newspaper article or a highly creative thing like a work of art?
-
Quantity and substantive use: How many copyrighted works do anyone want to use? Even if it's just a little bit, even if it's the “heart of work”, there may be no defensive conditions for reasonable use.
-
Impact on the market: Will that compete with the original author by using copyrighted works in the proposed way? What impact will this have on a larger market?
There are questions about every factor of fair use, AI, AI, Christian Mammen, an intellectual property attorney and managing partner at Womble, Bond and Dickinson Law Firm, told me in an interview. There is also a debate about whether the fair use factor applies to AI input, output or both. “Does this apply to the input aspect, to do the whole work in this training data, or to the output aspect, and may have an unrecognized, little impact in the output?” said Mom.
Tech companies are working on the exception to fair use because it can use copyrighted content without contacting all rights holders and paying licensing fees. For companies like OpenAI and Google that have already spent billions of dollars on development, a reasonable use exception will save a lot of time and money. Google said that fair use will allow it to continue to innovate rapidly. Openai took a parallel approach and said unhindered AI innovation is a national security issue.
Getting tech companies’ picky Blanche running AMOK with copyrighted content isn’t something creators are excited about. In March, more than 400 writers, actors and directors signed an open letter asking the Trump administration not to give Openai and Google a fairly used exception. Google and Openai are “fighting for special government exemptions so that despite their substantial income and available funds, they can freely take advantage of the creative and knowledge industry in the U.S. There is no reason to undermine or eliminate copyright protections that helped the U.S. thrive.”
The Copyright Office essentially pierced on the issue of fair use, and in a third report he said that sometimes fair use cases may occur, but sometimes it does not meet the necessary standards. Without federal legislation, we may have to wait for the ruling of some or all of these courts to set new legal precedents for copyright and fair use in the AI era.
What does all this mean for the future?
The copyright owner is currently in possession mode. But, in addition to legal and moral implications, copyright in the AI era raises important questions about the value of creative work, the cost of innovation, and the ways in which government intervention and protection are needed or should be conducted.
Mom said there are two different ways to view US intellectual property laws. The first is to create these laws to encourage and reward human thriving. Another is more economically concentrated; what we create has value, and we hope our economy can recognize this value accordingly.
“In most of our history, the humanistic approach and industrial policy approach have quite aligned,” Mammen said. But the generated AI emphasizes different approaches to copyright and IP.
“Does these laws exist primarily as problems with industrial economic policies, or do they exist as part of a humanist approach that values and encourages humanity to flourish by rewarding human creators?” asked Mom. “From the highest and most abstract level, I think this is one of the issues that these debates are forced.”
