Skip to the content
who owns ai-generated content
Article

Navigating the Complex World of AI Ownership and Ethics

2 October 2023

Jake Dunlap

“The million dollar question: Who owns AI-generated content?”

 

As almost every person and company embarks on this journey into the world of AI and AI-generated content, we’re starting to stumble upon questions about intellectual property, creativity, and ethical use.

Organizations are already using it to create mass amounts of content, written and visual, as well as to create strategies, write code, etc. It will only continue to become more ingrained in everything we do, especially in the professional world.

I recently had the pleasure of hosting a Modern Leader session with Keegan Caldwell, Attorney, Managing Member, and Founder at Caldwell Law, in which you’ll hear that many of these questions are new and unanswered or a grey area. Still, things are going on today that will shape how we deal with these questions in the future. Particularly around

    • Copyright vs. Patents (when it comes to AI)
    • ChatGPT Content Ownership
    • Content Ownership & Plagiarism
    • Ethical Use & Misinformation

I’ll summarize the key points Keegan and I discussed live here, so I highly recommend watching the recording if you have time for a more comprehensive understanding and example cases thus far.

*Disclaimer: I have to say it. The information provided in this article is not legal advice. Always consult with a legal professional regarding any intellectual property concerns.

 

Recap: AI & Intellectual Property – Who Owns It And What This Means For The Future

Skip ahead to:

    • 7:46 – The million-dollar question: how does ownership of AI-generated content work?
    • 9:53 – Can a non-human author be an author and under copyright protection?
    • 12:15 – Patents protect innovations. In this case, people almost certainly get patents on products that use AI to create them all the time.
    • 14:10 – Governments are set up to protect citizens, not machines. It would be a slippery slope.
    • 14:40 – Who owns the content generated by ChatGPT? OpenAI’s policy states that the content generated and the content input assigns all the right title and interest to the user. This means users can use the content for any purpose, including commercial ones.
    • 16:00 – Open AI doesn’t use your inputs for training data.
    • 19:00 – The fear is that if we’re inputting our information, another company somewhere is benefiting from it, but based on their policy, this information isn’t shared or used to improve ChatGPT.
    • 20:55 – What issues could arise around infringement or usage of content when creating content from generative AI? So many companies are already cranking out content using generative AI.
    • 26:00 – Impacts AI is going to have on human creativity and originality. I’m guilty of initiating new project ideas from a prompt, etc.
    • 28:00 – A significant point Keegan makes here is the accidental amplification of biases in language models because of the training data it’s served.
    • 31:00 – AI may take over some current jobs, but it will open up people to work on other things. Think of AI as repositioning or changing the jobs we have today.

 

Copyright vs Patents

How does ownership of AI-generated content work? This is where we start the conversation when trying to understand who owns what. We talk primarily about intellectual property and copyright in this session, but it’s important to understand the difference between the two because there are cases to be made using AI for both.

Copyright pertains to the expression of an idea, such as written works, music, and art. In the AI world, the question is, if you input something into ChatGPT or similar, is that copyrightable? The short answer is no, and the argument was there must be a human author.

On the other hand, patents protect inventions or processes for a certain period, ensuring exclusive rights to the inventor. Today, people get patents on products that are produced using a certain level of AI all the time, but a human engineer typically designs the product, and AI is used to optimize it. 

Keegan references several cases shaping the conversation around AI copyright and patent terms. One that stands out is Steven Thaler, who is a creator of artificial intelligence products and sued for both copyright and patent protection with AI as the author and inventor. The case was rejected.

So, at this point, we can assume there has to be a human author or inventor or at least meet a certain threshold for work to be legally covered.

 

ChatGPT Content Ownership

OpenAI, the organization behind ChatGPT, has a clear stance on ownership. They assert that the content generated and the content input assigns all rights, title, and interest to the user, “but that the use of the model is subject to the terms and conditions set by OpenAI.” This means users can freely use the content for any purpose, including commercial ones. 

 

Content Ownership & Plagiarism

This is an important one for people. Especially after OpenAI/ChatGPT says you own the content you input and the output, BUT ChatGPT and other training models are basing their answers off information it was trained on – possibly pulling from the web and other copyrighted places.

My question to Keegan was how this would be handled legally if the issue of infringement or usage of content on something created by AI came up. He broke it down into a couple of different scenarios: 

Plagiarism of AI-Generated Work: If someone copies content produced by AI, the determining factor is what portion, if any, of the AI-generated work can be copyrighted. Which we talked about in the first section, is hard to do.

Infringement by AI-Generated Content: It’s crucial to determine what was copied. Was it copyrightable content or generic information that falls outside of copyright? 

For example, in a telephone book court case, where a defendant copied the contents of a phone book, the court ruled there was no infringement because what was copied was not copyrightable material – it was just factual.

But as a second possibility, what if an AI-generated work uses copyrighted material as training data? In this case, is the output derivative work? There are pending lawsuits on this one, so to-be-determined.

 

Ethical Use & Misinformation

Aside from the issues around copyright, it was vital that we touched on other ethical impacts around originality and creativity. 

What came out of this was possible issues around how LLMs are trained to give certain outputs based on biased training data, misinformation in outputs, and the takeover of jobs.

Of course, none of these things were intended or the goal of AI, but we can’t ignore how it’s already shaping our world or change that it’s happening, but we can determine how we react to it.

On this first topic, it was interesting to think about the potential of biases even in AI training data, which can inadvertently lead to skewed results because the model isn’t being fed an equal amount of objective or varied information.

You can go ask ChatGPT right now if it’s biased toward certain groups, and it will tell you it tries to be as neutral and unbiased as possible, but there are a few things to consider, etc.

In terms of misinformation, if you use ChatGPT, you know, sometimes it’s just factually wrong, but misinformation from other types of AI can also be very deceptive. Keegan discusses two types of misinformation in the context of AI:

Misinformation through Reasoning: This occurs when the AI model attempts to sway people to a point of view through factual assertions. 

Misinformation through Rhetoric: Here, the AI might produce content disguised as a well-reasoned argument but appeals to biases, emotions, or preconceived notions to sway opinions.

Lastly, the landscape of professional fields is undergoing a significant shift due to the rise of AI and automation. While there’s a prevalent fear that AI might replace certain jobs, viewing this change from a broader perspective is essential.

This change is not about job elimination but job evolution. As specific roles become automated, it paves the way for individuals to focus on more creative and strategic tasks. While tools and AI can automate repetitive tasks, the human element – creativity and personal touch – remains irreplaceable… for now.