Your Writing and Your Art Might Be Owned by Public Domain

I can steal your AI-generated intellectual property. Current facts from creative experts and stakeholders

The storm of controversy surrounding AI continues to roll in. (photo by author)

Copyrights for AI-generated writing and imaging have been under deep scrutiny for more than a year. The battle is on; the storm rages. If you are a writer or a self-publisher, or you use artificial intelligence in your microbusiness or your solopreneur gig, it’s essential that you understand where this is going. 

It’s complicated and intimidating. I've attended and followed Copyright Alliance webinars and bulletins since AI-generated content slammed onto the publishing scene. I hope to give you some helpful insight into where this all is leading.

Do humans control AI generation output, or do machines?

In one seminar, I heard a pointed opinion from Chris Callison Burch, Associate Professor at the University of Pennsylvania. He specializes in natural language processing‬ and artificial intelligence. Burch declared that creative works produced with AI should be copyrightable because human users alter the content in non-trivial ways. He believes the US Copyright Office underestimates the amount of input humans add.

“If the Office fails to define training AI with literary works as fair use, it would make training these bots almost impossible.” He explained that such trainings are based on random input. “To get an output, you need a prompt, a random seed, and a model,” he said, so the office could require proof of those items in order to qualify for copyright registration. Determining fair use on a case-by-case basis would restrict the development of these technologies. So fair use should be a standard definition."

But is that truth or opinion? The world-wide discussion about copyrights for AI-generated writing and imaging is still a contentious debate, though many countries have decided not to allow registration of copyrights for non-human-created material.

In the United States, most artificial intelligence (AI)-generated content is typically in the public domain and not copyrightable. This is because the U.S. Copyright Office requires human authorship for copyright protection. However, the companies that own the AI-generating platforms may reserve certain rights in their terms of use and may essentially own anything made on their platform. Read those terms. 

Works that contain AI-generated material may be copyrightable if they meet certain conditions:

• The work includes human arrangements that are sufficiently creative (a subjective parameter)

• The work combines AI-generated and human-authored material (difficult to prove)

If a work is determined to meet these conditions, the author may be able to claim copyright protection for their own contributions. They must also identify and disclaim the AI-generated parts of the work when registering the copyright. 

There are similar rules in Europe; a recent court decision said, "As AI is not a natural person, it cannot be the creator of a work under European copyright law." China has disagreed and awarded copyrights to at least one completely AI-generated intellectual property.   

How do writers see this problem?

Image by Fathromi Ramdlon from Pixabay

Mary Rasenberger, Executive Director of the Authors Guild and Authors Guild Foundation, pointed out what platforms writers are using, and many writers definitely are using AI platforms. She said, “GPT 4 is most common. Bard gets 8% of users.”

In early 2023, about 30% of writers admitted to using ChatGPT or its ilk for brainstorming — plot, article ideas, character types, structure, and organizing. One-quarter of the use is for marketing and communications like email, flyers, and brochures. At least half of all writers use something like this for proofreading. While only 7% of writers interviewed admitted to fully generating content, 90% say 10% or less of their writing is generated. Most writers surveyed said their personal touch, voice, and additions are very important to the quality of their product.

In 2024, according to DDIY.co, the picture shifted dramatically.

1. 48% of businesses and organizations use some type of ML (Machine Learning) or AI

2. 58% of companies that use generative AI use it for content creation

3. Bloggers who use AI spend about 30% less time writing a blog post

4. The global AI market is worth approximately $196.63 billion in 2024.

And 88% of those who earn money by writing online say they use AI to generate that content, at least in part. A large number of writers have convinced themselves that AI writing is a viable path to selling content.

Rasenberger said she is personally aware of growing numbers of writers losing business or finding it more difficult to sell their writing since the 2022 release of AI-generators. “One writer,” she said, “lost 75% of her income due to AI.”

Generative AI can’t think or express feelings,” Rasenberger went on, “it can only combine what it has learned. Researchers report that no one knows exactly what the training data is, but it pretty much includes every published book. Authors should be able to choose if their work is used to train AI.”

Training AI generators—fair use or theft?

Though developers are pretty cagey about disclosing exact information, most articles online and other works on the internet have been copied and used to train bots. The Authors Guild wants to see the development of collective licensing to allow developers to use data only if they compensate and/or credit authors. 

Edward Hasbrouck of the National Writers Union chimed in, “Without training material, prompts, and human contribution, AI would turn out garbage. GIGO applies — garbage in, garbage out. The works used to train AI are most likely publicly accessible web content.” 

Keep in mind that publicly accessible does not automatically mean public domain. Hasbrouck suggested that in the decades since the internet became a thing, the copyright office has never adequately addressed policies to protect web content. It’s all just out there, and most of it is copied and used at will, but that doesn't make it right or legal to do so.

Copyrights for AI-generated material not made simple

The conversation has grown to encompass programming and coding, processes also shifting from all-human to partially AI-generated. Jule Sigall is Associate General Counsel for Copyright in Microsoft’s Legal & Corporate Affairs. Sigall says, “The most popular tool in coding is Co-Pilot. It helps developers bridge from one language to another that they may not know as well.”

Nobody knows who is telling the truth and who isn't

But according to Leigh Hennig, author, editor, and engineer, one hundred editors and publishers interviewed find that authors don’t alter content as frequently as we may think. Many times, a user cobbles together a vague prompt, receives output, scrapes it from the platform, and sells or publishes it exactly as it came out of the bot.

The seesaw of copyrights for AI-generated writing and imaging

One webinar participant was adamant that the Copyright Office should not engage in investigations of whether there is sufficient human involvement in a creative work. There are too many variables, definitions, and methods for anyone to be able to set boundaries. 

Others maintain that incentives for human authorship must be protected. Did the human create the actual expression of thought in consideration? Is that the decider? And what percent of generated material should require an author disclaimer?

Soon, nearly everyone will be using this technology, which may seriously alter the definition of authorship or written work. We may, said one speaker, end up with rules for the future, not for the past. 

It seems like that has been the end game with all new technology. Technology proponents always want new technology to be given every advantage. After all, there is always money at stake. 

Additionally, workers who worry about losing their jobs to machines typically want stricter regulations so they can compete in the labor market. Now, writers and authors are concerned about the authorities invalidating or canceling their copyright registrations if someone determines that the intellectual property is less than human-created. It isn't easy to 100% detect whether writing is AI-generated or human-created.

At present, a writer does not own their content if it is determined to have too much of an AI component. Whether or not we label our content "all rights reserved," the rights may be in jeopardy if we allow a bot to assist in our writing. Is it worth the risk to save an hour of writing time?

I'm a huge proponent of humans creating the body of written material that's important to our culture. If a bot creates marketing mambo jumbo or encyclopedic entries, I'm not too troubled. But what if machines upchuck our future legacy of tales, histories, poems, songs, and sagas and the record of the human experience becomes irrelevant?