Discover the new KeyShot

Design for
tomorrow.
Market today.

Discover the new KeyShot

Design for
tomorrow.
Market today.

Copyright law and AI: How to make sure your images don’t infringe copyright laws

Categories

Date

October 10, 2023

Share This Story

Written By KeyShot
The KeyShot crew fills you in with the latest KeyShot tips and tricks, insight into 3D rendering technology and the people creating the coolest visuals across the engineering, product design and entertainment industries.

2023 was undoubtedly the year that AI went mainstream

Since Chat GPT launched on 30th November 2022, the endless possibilities of generative AI have captured the world’s attention like nothing else.

In fact, Chat GPT caught on so quickly that it set the record as the fastest-growing consumer app of all time. 

Thanks to this newfound global enthusiasm for using generative AI tools to create text, images, video, and other content, the sector is expected to be worth almost $45bn in 2023, and with a compound annual growth of 24.40%, is expected to be worth a whopping $207bn by 2030.

With all this growth and constant innovation, it’s safe to say that generative AI is not going anywhere and is set to be a critical part of business workflows everywhere.

As exciting as this new frontier of content creation is, it’s important you keep up to date with the potential legal implications of using generative AI in your organization – to make sure you don’t infringe copyright law.

So, we’ve put together the ultimate guide to copyright law and AI – including input from legal experts in the US and UK – to help you manage your AI-generated digital assets with ease.

  • Table of Contents:

    • Using generative AI to produce images

    • The evolution of AI-generated images 

    • Issues with generative AI images

    • Does copyright law protect AI-generated images?

    • Does copyright law protect images from AI training?

    • Who owns an AI-generated image?

    • International law and AI copyright

    • Copyright best practices for generative AI images

    • Conclusion

Using generative AI to produce images

Of course, generative AI isn’t limited to the text-based outputs of tools like Chat-GPT or even Google’s competitor product, Bard, which was launched in March 2023. 

Generative AI is capable of producing a wide variety of media content, such as:

> Voiceovers in any accent or language

> Music in any style or genre 

> Computer code in any language

> Realistic video, and special effects.

However, one of the key advancements in AI technology is the limitless possibilities it offers when it comes to image generation.

Using tools like Midjourney, as well as countless integrations into existing tools like Photoshop, Canva, and Shutterstock, means anyone can quickly leverage the incredible power of generative AI in their content workflows.

However, with great power comes great responsibility – so it’s important to understand how you and your company can use generative AI effectively to produce images and other content without the risk of infringing copyright laws.

The Evolution of AI-Generated Images 

The AI landscape is changing all the time – so you need to stay on top of the latest developments to make sure you don’t get left behind. 

Generative AI has gone from science fiction to a mainstream media production tool in what seems like no time at all. 

To give you an idea of how quickly things have evolved, we’ve put together a brief timeline of developments over the past 6 years. 

  • The evolution of AI-generated images

    • 2018 NVIDIA develops StyleGAN – a machine learning model to offer the creation of detailed, high-resolution images. 

    • 2019 – Competitive machine learning models (GANs) are developed by researchers to generate medical images to help improve diagnosis and treatment planning. 

    • 2020NVIDIA releases StyleGAN2 – this new and improved version allows for even more realistic image generation.

    • 2021OpenAI introduces DALL-E, which allows end users to generate contextual images from simple text descriptions.

    • 2022Midjourney is launched in March 2022, creating a more accessible way for end users to produce AI-generated images. 

    • 2023 – Generative AI for images is now a common feature in almost every tool or app used in content production workflows, e.g., Photoshop, Canva, and Shutterstock.

Issues with Generative AI images

However, almost as fast as the technology is advancing, questions are being asked about the legal impact of using generative AI.

At the time of writing, generative AI has brought the Hollywood movie industry to a total standstill as part of the Screen Actors Guild strikes. 

One of the key concerns of the Hollywood strikers is the potential for movie studios to replace scriptwriters with AI tools like ChatGPT. 

Compounding the issue further, there is huge concern about these generative AI tools being trained on the previous work of writers without any form of permission granted or compensation offered – essentially using their IP for free.

On top of this, Hollywood actors are worried that AI replicas of their likeness could be used in movies without their permission or financial compensation – impacting their legal Personality rights.  

These AI replicas – also known as Deep Fakes – also pose a more sinister threat, where fake videos featuring people’s likenesses can be created without their knowledge or consent to appear in any number of unsavory scenarios.

A recent example saw podcaster Joe Rogan appearing to promote a particular coffee brand (he didn’t), but as part of a social media campaign, it appeared very convincing. 

So, while the advancement in AI technology creates amazing opportunities for improving the way people efficiently produce images and content, It also raises questions.

Do traditional copyright laws, intellectual property laws, and personality rights apply in the world of generative AI?

Let’s find out. 

Copyright law and AI-generated images

Does copyright law protect AI-generated images?

** Disclaimer – Before we get started, it goes without saying that anything in this article, including input from legal professionals we have interviewed, does not constitute legal advice and should be used for informational purposes only. **

The application of traditional copyright law to AI-generated content is a subject of much debate within the legal world right now. 

This is because the traditional copyright laws were written long before artificial intelligence was developed.

This means that many of the legal frameworks still in use today around the world are all based on the concept of human authorship and ownership.

Put simply, copyright law usually requires original content to be created by a human in order for it to qualify for copyright protection against being misused by another human.

This is where the issue with AI arises, as the content isn’t generated by a human.

So, you’re probably now asking yourself two questions:

Is AI-generated content protectable under current copyright law?

Is your original content protected by copyright law if it’s used to train AI models? 

The answer, of course, is complicated. 

Copyright law is slightly different in every country, but if we take the US as an example, their laws say that original works created solely by machines without any direct creative input from humans are not protected by copyright law.

This means that while the underlying AI technology that creates the imagery is likely to be protected by copyright and patent laws, the images and content it creates are not. 

Until the current legal frameworks are updated, it means that imagery and content created by generative AI tools seem to exist in a little bit of a legal limbo.

But what about the other way around? Does the law protect original content from being used to train AI technology?

Does copyright law protect images from AI training?

In 2023, there have been several notable lawsuits brought against generative AI companies.

The key premise behind these cases is that human-created works protected by copyright law are being used – without permission or compensation – to train generative AI models.

  • These cases include:

    • Mona Awad and Paul Tremblay vs. OpenAI: The authors filed a lawsuit against OpenAI, claiming that ChatGPT can allegedly generate accurate summaries of their works – which may indicate the model was trained on their copyrighted material.

    • Sarah Silverman vs. OpenAI and Meta: Comedian Sarah Silverman is suing both companies for allegedly using her memoir, “The Bedwetter,” to train their AI models without her permission. 

    • GitHub, Microsoft, and OpenAI: A class-action suit was filed against these companies over GitHub’s Copilot tool. The suit alleges that Copilot’s AI copies and republishes code from GitHub without proper attribution to the original authors.

    • Stability AI, Midjourney, and DeviantArt: These companies are facing a lawsuit for allegedly infringing artists’ copyrights by training AI on their works. The claim says that the AI could then produce unauthorized images in the same style as a result. 

    • Getty Images vs. Stability AI: Getty filed a complaint against Stability AI for allegedly using millions of images and associated metadata owned by Getty to train their AI models.

Not all these cases specifically cover image rights, but they help to highlight the broader context of potential copyright infringement with AI.

Who Owns an AI-Generated Image? 

One of the key questions around AI content copyright is – Who owns it?

The answer is not as obvious as it may seem, as there are potentially several options:

> Is it the developer of the AI software?

> The user who used the AI tool?

> The AI itself?

To get to the bottom of this, we asked two IP-focused legal experts for their take, one from the UK and one from the US:

Mark MacleanStraightline Legal Ltd

“The UK is one of the few countries to protect works generated by a computer where there is no human creator. The “author” of a “computer-generated work” (CGW) is defined as “the person by whom the arrangements necessary for the creation of the work are undertaken”.  

It is highly unlikely that the computer program could own copyright in the CGW – the owner must typically be a human or a corporate entity. It is important to note that the terms of use for a particular AI platform may contractually allocate who owns the image and associated copyright (should copyright subsist), and that will be the starting point in many cases”.

 Sekou Campbell – partner at Culhane Meadows  

“These are critical unanswered questions. We now have clues as to how certain stakeholders may answer these questions. The US Copyright Office, for instance, has issued guidance referencing some early decisions on certain materials presented for copyright registration. While the US Copyright Office’s decisions are persuasive, these questions remain unresolved as a matter of law. 

With that caveat, the US Copyright Office has taken the consistent position that AI-generated content challenges the “human authorship” requirement under statutory and precedential case law. “For example, when an AI receives solely a prompt from a human and produces complex written, visual, or musical works in response, the ‘traditional elements of authorship’ are determined and executed by the technology—not the human user.” 

From speaking with Mark and Sekou, our takeaway is that the question of ownership of the image is complex. 

The law in both the US and the UK is starting to move away from the limitations of the old-fashioned human authorship criteria to better understand that a computer or AI model is capable of creating an image or other copyrightable content.

However, if you have a suitable contract of usage agreement in place with the AI model you’re using, then you should be able to claim ownership of the image, even if you can’t claim creation rights. 

International Law and AI Copyright 

Ultimately, though, the issue of copyright infringement through AI is a universal one that affects content creators of all types – in all countries.  

However, as with all laws, there is no uniform global standard – every country does it slightly differently. 

For example, the EU proposed the Artificial Intelligence Act in 2021, which serves as a pre-emptive ‘catch-all’ for all types of automated technology, not just AI itself. The key parts of the act aim to regulate the use of these technologies in sectors like Healthcare whilst prohibiting their use in ways that can threaten livelihoods. 

Meanwhile, the United States has yet to pass any specific legislation at a national level, instead offering federal guidelines known as the AI Bill of Rights. Instead, individual states are being tasked with regulating AI within their own jurisdictions.  

China does not yet have laws in place around AI copyright. Instead, they have set a target for their private AI industry to achieve $154bn of annual revenue by 2030

So, to be on the safe side, regardless of which country you’re in or what kind of generative AI you’re using, you’re going to want some universal best practice tips to follow that will be applicable anywhere. 

Luckily, we’ve got you covered.

Copyright Best Practices for Generative AI Images 

So you can make sure that you’re not infringing copyright law with your AI-generated images, there are a couple of simple best practice tips you should follow: 

  1. Do you have a commercial license agreement in place? 
    Regardless of which AI tool you use to generate your images or other content, you need to make sure that you have a suitable license in place to use the content in a commercial environment. Some AI tools have been trained on copyrighted data under the guise of ‘research’, and the content they produce is not meant to be used in a commercial environment. Check the license/contract /terms and conditions regarding commercial terms for the tools you’re using and document this evidence. 

  2. Monitor all output 
    Sometimes, there is just no substitute for having a good look at the images being created. Do they look like something you’ve seen before? What happens if you do a reverse Google image search on them? It’s better to spend a little time checking your images, than end up facing a lawsuit for copyright infringement. 

  3. Document everything 
    Using the metadata tagging options in your Digital Asset Management system – be sure to document when the image was generated, what tools you used, and any other relevant information to help you prove the image was AI generated and not intended to be an infringement of anyone’s copyright. 

  4. Stay updated on legislation
    In much the same way as we’ll have to revisit and update this article regularly, you should also keep yourself up to date with AI legislation in your jurisdiction. The landscape is regularly changing and evolving quickly, so you need to keep pace with it to avoid potential pitfalls in the future. 

By focusing on these 4 simple best practice tips, you’ll be well on your way to mitigating any potential copyright infringement risk that may occur with the use of generative AI images or content. 

Conclusion 

We’ve talked in depth about copyright law and generative AI images – as well as other AI-derived content like text, music, video, and voice audio, which face the same issues. 

If there’s one thing you should take away from this article, it’s that the topic of copyright law and generative AI is continually evolving.

Because new lawsuits are starting all the time and individual countries are beginning to change their stance on AI copyright, it means that the landscape could look very different in a few years.  

Ultimately, though, if you are using – or planning to use – AI-generated content as part of your digital asset mix, then it’s incredibly important for you to stay up to date on the potential copyright infringement risks involved. 

Want to learn more about how to manage all your digital assets – including AI-generated images – with a digital asset management system?

Find out more or book a demo

FAQs 

Q. Can artificial intelligence hold copyright?

A. In most countries, AI-generated works are generally not eligible for copyright protection, which requires human authorship. However, in the UK, they may be able to qualify as computer-generated works (CGW) – which are afforded some protection under copyright law. It is worth noting that the legal landscape around AI is constantly evolving, and some countries are considering extending copyright laws to include AI-generated content. 

Q. How does artificial intelligence affect copyright law?

A. The widespread adoption of AI has introduced new complexities to be considered in relation to copyright law, especially around issues of data scraping and training and the rights (or otherwise) to access copyrighted material in order to do so. Traditional copyright laws were not designed with AI in mind, as a result, there are ongoing efforts to consider how best to update copyright legislation.

Q. Who owns the copyright of an AI image?

A. The law in both the US and the UK is starting to move away from the limitations of the old-fashioned human authorship criteria to better understand that a computer or AI model is capable of creating an image or other copyrightable content.

However, if you have a suitable contract of usage agreement in place with the AI model you’re using, then you should be able to claim ownership of the image, even if you can’t claim creation rights. 

Q. Does AI training violate copyright?

A. AI training can potentially violate copyright laws if the data used for training includes copyrighted material that was obtained without permission. As we discussed in the article, this has led to several ongoing lawsuits against AI companies, claiming that their models were trained on copyrighted works – the outcomes of which could set important precedents.

Q. Can you train AI with copyrighted images?

A. Training AI with copyrighted images can potentially be considered legal under the right circumstances if it is done for research purposes. However, where the AI is trained with commercial intent, or the outcomes of the training are made available in a commercial product, this is likely to be considered a violation of copyright laws.

Related Posts

Why 3D and CGI Improve Ecommerce Sales and Customer Retention

Why 3D and CGI Improve Ecommerce Sales and Customer Retention

How to Choose the Right Digital Asset Management System 

How to Choose the Right Digital Asset Management System