[Activity insight] Who owns AI-generated content?
With generative AI tools, we have all become (or have the potential to become) creators of all kinds of media. But do we own what we have generated? Can we monetize these products?
Last week, I was invited to give a short talk at the SIG Morning event of BRAZ-TESOL, where I was talking about how language teachers can use GenAI in their lesson planning process and to create materials. At the end, I asked a thought provoking question:
What’s going to happen to professional materials developers (and everyone else in the publishing industry) if every language teacher can now design, use, and publish their own more tailored and more up-to-date materials more quickly?
The question is indeed a rhetorical one because it’s hard to say. The fact that now everyone is capable of designing their own activities might mean that course books become less of a go-to choice (to be honest, I don’t remember the last time I used a course book for teaching but it’s true that I mostly do corporate classes or one-to-ones).
Also, I guess the more tailored and more up-to-date your materials are, the less you’re going to want to buy published materials. But I think in institutional settings, books will remain a staple for longer to provide consistency and reliability to the stakeholders (parents, students, clients).
But my real proposition is that publishers will have to step up their game to stay in business, because the competition might not come in the form of self-published EFL teachers but in the form of declining demand. The bigger the company, the slower it can adapt to changes and the less likely it can include more diverse topics.
Who owns what in the world of AI-generated content?
Still, if we take freelance teachers designing their own materials and thinking about copyrighting and selling them, can this actually be done legally?
I looked at the Terms of Use of several AI generators, and this is what I’ve found. It’s definitely not a complete list but we can still draw some conclusions from it.
Text generators
I checked OpenAI and Anthropic, and regarding user input, they say that the user should make sure they own the rights to all the content they upload into the chatbots. This means that it’s your responsibility to make sure that what you use as input is entirely yours. So, pay attention not to upload or type in any already published or copyrighted materials.
Regarding output, both companies say that if you co-created something with ChatGPT or Claude, you are entitled to all the rights, titles and interest that come with the generated product. In other words, you’re free to publish, own copyright to, and sell your creations as long as they are the result of a joint effort with a chatbot (it’s unclear how much of your input actually makes it unequivocally yours). This also means that if you ask for something to be created “in the style of…”, it’s not yours.
Check my conversation about this with ChatGPT here.
Image generators
Again, regarding input, Midjourney and Tengr say that you need to make sure you upload content that’s legally yours or royalty free. You also need to guarantee that you don’t upload anything that violates copyright or trademark rights or ethical considerations.
The output is entirely yours. However, Midjourney states that “if you are a company or any employee of a company with more than $1,000,000 USD a year in revenue, you must be subscribed to a “Pro” or “Mega” plan to own Your Assets.”
Song generators
While song generators, Suno and Udio, give you ownership rights (including commercial use rights) and don’t even require you to credit them if you’re on a paid plan, it’s even more complicated and controversial to decide whether the generated products are entirely unique and belong to you. But more on this later.
These generators also let you upload a sample of your own voice, which you can then use for creating songs that sound just like you. But what you need to bear in mind is that once you uploaded your voice, it belongs to the companies.
Stay safe, stay transparent
The best approach to dealing with AI-generated material is to be transparent about its use. You may want to use the following templates for this:
Created with Midjourney based on prompt XYZ.
This resource was co-created with the assistance of an AI language model (ChatGPT) and adapted by the author for English language teaching purposes.
Some elements in this worksheet were developed with help from ChatGPT and refined by the teacher for classroom use.
Artists’ fight against scraping and data mining
There have been many legal cases against various companies owning generative AI tools. Artists rightfully believe that training models on their creative works constitutes copyright infringement and, therefore, companies should either stop this activity completely or reimburse artists for all the financial damage they have caused.
The problem is that, again, nothing is black and white. Two recent cases (Bartz v. Anthropic PBC (Judge Alsup, June 23, 2025) and Kadrey v. Meta Platforms, Inc. (Judge Chhabria, June 25, 2025)) show that what seems to matter the most is whether the use of training data is transformative enough. If it is, as it was how the judges ruled in both cases, then the companies showed fair use and no copyright infringement happened. The models don’t memorize the training data in order to produce the exact same result but they learn from these texts to produce something new. I’m not sure how consoling this is to writers but I do hope that LLMs won’t be able to produce texts that are hauntingly similar to real authors’ works.
The picture is even more complicated in the case of images, videos, and audio because it’s much more obvious if similar or the same product has been generated. And there are still many ongoing cases without a clear ruling or decision. What seems to be the direction forward is to include opt-out schemes that artists can use to prevent their work from being used in training. Whether this is going to be legally binding is an interesting question.
What artists have been trying to do lately is “poison” their works. During data poisoning, “the attacker introduces misleading information, modifies existing data, or deletes important data points. The goal of the attacker is to mislead the AI into making incorrect predictions or decisions” (source). Or they can also edit the metadata to try and protect their works from copyright infringement.
Activities in our book
We’ve got several activities that deal with the gray area of ownership. Since the question is definitely not easy to answer, these activities include lots of discussion and debate.
Plagiarism or not?
Students need clear guidance on using AI tools responsibly for schoolwork without unintentionally cheating. In this activity, students review a list of possible ways to use AI in their studies and decide which are acceptable and which aren’t. This is followed by a class discussion on ethical use, fairness, and originality.
Who owns it?
This activity encourages students to think critically about the consequences of using AI to generate ‘creative products’ such images, music, or text. By engaging in debate, students will explore important concepts such as copyright, intellectual property, and what should be available for everyone to use for free.
What could go wrong if…?
Sometimes, it is hard to realise how small actions, like uploading private or copyrighted documents online, can lead to illegal or harmful misuse. This activity helps students understand the risks of uploading fi les that shouldn’t be shared or asking unethical or unsafe questions to an AI chatbot, such as how to print money.