Does AI Violate Copyright Law? | A Conversation with Copyright Lawyers
From @curiousrefuge Copyright Lawyers Kristyn Webb and Julie Greenberg. This is an interesting discussion about the current (jul, 2024) state of AI as it relates to Copyright laws. This video is one hour and encourage all artists to at least listen to it once. Just turn it on while you are working on something. But, if that is TLDW, I took some notes below. My notes are paraphrased from the video.
In this video, we hear from Kristyn Webb and Julie Greenberg, two IP and Copyright lawyers who share their thoughts on the legal side of using AI for creative work.
We are very excited to share with you our most recent office hours with Copyright Lawyers Kristyn Webb and Julie Greenberg. This is such an insightful conversation and we are sooo grateful for their willingness to chat with our community and share about the current state of AI as it relates to Copyright laws.
© All rights reserved. Contact jgreenberg@fishsteip.com for permissions. Not intended as legal advice.
Helpful Links:
What is Copyright?
Copyright is a legal protection for creative works, automatically granted upon creation. While displaying the copyright symbol (©) is no longer mandatory, it is still recommended. Copyright typically lasts for the life of the creator plus 70 years.
Not Copyrightable
- Ideas
- Facts
- Concepts
- Typefaces
- Names
- Titles
- Lists of ingredients
- Recipes
- Phone books
- Fashion (utilitarian aspects)
Exclusive Rights
Copyright grants the creator several exclusive rights:
-
- Reproduction: The right to copy the work.
- Derivative Works: The right to create adaptations, such as translations, movie adaptations, or sequels. (Note: Fan fiction and spoofs are not typically covered.)
- Distribution: The right to distribute copies of the work.
- Public Display: The right to display the work publicly.
- Public Performance: The right to perform the work publicly.
- Digital Audio Transmission: The right to perform the work via digital audio.
Originality
For a work to be copyrightable, it must possess minimal creativity and be original to the author. This excludes common shapes and symbols, requiring only a very modest level of creativity. The work must be independently created, not copied from another job.
The concept of originality implies that the work originates from the author. Even if two works appear similar, each creator retains copyright over their respective creations. For instance, if a homeowner collaborates with an architect, the architect owns the copyright to the architectural design, despite the homeowner’s direction. The person who executes the creation is the copyright owner.
Naruto V. Slater, 888 F.3d 418 (9th Cir. 2018) Case Study
This case highlights the crucial question of copyright ownership. A freelance photographer, David Slater, aimed to photograph monkeys in a specific rainforest. Unable to capture satisfactory images himself, he set up cameras with remote shutters.
Since Slater did not press the shutter button to take the photos, the question arose: who owns the copyright?
When Wikipedia published the photos, Slater sued, claiming ownership of the copyright. The court examined the case and asked, “Who pressed the shutter?” It was revealed that the monkeys had pressed the shutter in each instance. Consequently, the court ruled that Slater did not own the copyright because he did not create the work. According to U.S. law, only a human can hold copyright, meaning the monkeys’ photos could not be owned by Slater or the monkeys.
This case underscores a fundamental principle in U.S. copyright law: a human must create the work. This point is especially relevant in discussions about AI and copyright.
U.S. Copyright Basics
- Independent Contractors: Own the copyright to their works.
- Copyright Transfers: Only occur through a written assignment.
- Licenses (Permissions): Can be granted verbally.
1989 Supreme Court Decision: Community for Creative Non-Violence v. Reid
In 1989, the Supreme Court issued a landmark decision in Community for Creative Non-Violence v. Reid, fundamentally altering the understanding of copyright law.
Key Points:
- Employee vs. Independent Contractor: The Court ruled that only if a person is an actual employee (not an independent contractor) of the entity that paid them to create the work does the paying company own the copyright.
- Artists’ Ownership: As a result, artists retain ownership of their work unless they are legitimate, fully employed employees of the entity that commissioned it.
- Written Assignment: There must be a written assignment to transfer copyright ownership from the artist to the commissioning party.
To summarize, the 1989 Supreme Court Decision established that, unless specified in a written agreement, the creator of a work retains the copyright, regardless of who funded the creation.
Copyright Registration
Copyright is automatically granted upon creating a work; you do not need to take further action to establish it. However, for enhanced protection, it is advisable to register your copyright with the Copyright Office.
Importance of Registration
- Legal Protection: Registering your copyright provides more substantial legal rights. With registration, you can file a lawsuit for infringement once you obtain the registration.
- Court Access: Without registration, you cannot pursue legal action for copyright infringement in court. If someone unlawfully copies your work, you need the registration to be heard in court.
- Proving Damages: Without registration, you must confirm the extent of your damages in court, which is challenging in art.
Statutory Damages
- Pre-Registration Protection: If you have registered your copyright before the infringement occurs, you are entitled to statutory damages.
- Statutory Damages: These damages are predetermined amounts that do not require proof of actual harm.
- Innocent Infringement: Up to $33,000 in statutory damages, even if the infringer was unaware of the copyright.
- Willful Infringement: Up to $150,000 per infringement if the infringer knew about the copyright and copied the work regardless.
Registering your copyright can provide significant financial protection and legal leverage in the event of infringement.
Infringement
- No Defense: There are no acceptable defenses for copyright infringement.
- Single Copy: Infringement applies even if only one copy is made.
- Social Media Posts: Posting copyrighted material on social media still constitutes infringement.
- Attribution: Giving credit to the original creator does not exempt you from infringement.
How Does Copyright Handle New Technologies?
With the advent of AI technologies such as:
- Chat GPT (text)
- Stable Diffusion/Diffus (images)
- Midjourney (images)
- Sythesia (video)
- Udio (music)
- RAGHAV (images)
- Dall-E (images)
significant questions arise about copyright implications.
Key Issues
1. Inputs
AI systems are trained on massive datasets, often sourced from the internet without the original creators’ permission or compensation. This practice raises concerns about copyright infringement and the loss of control that creators should have under copyright law.
2. Outputs
Determining the copyright status of AI-generated outputs is complex. Questions include:
- Do creators have any copyright interest in AI-generated works?
- Are there liabilities associated with using or distributing these AI-generated works?
Example Case Study: VCRs
When VCRs emerged, users began recording TV episodes to watch later, creating copies of copyrighted works. Lawsuits argued that this practice was unauthorized copying. However, the courts ruled that “time-shifting” was not infringement, showing a protectionist stance towards new technology. This precedent highlights the evolving nature of copyright law in response to technological advancements.
Other Examples
- Google Books: Faced legal challenges for digitizing books without permission.
- Napster: Peer-to-peer music file sharing was ruled as infringement due to public distribution and local copying, illustrating the impact of new technologies on copyright law.
As AI continues to evolve, copyright law will need to address both the use of existing works for training and the status of AI-generated outputs to ensure fair compensation and control for original creators.
A.I. Lawsuits (July 2024)
There are currently 20 pending lawsuits related to A.I. and copyright infringement:
- Alter v. OpenAI (S.D.N.Y.) – Authors (Input Lawsuits)
- Andersen v. Stability AI (class action – N.D. Cal.) – Visual Arts (Input Lawsuits)
- Concord Music Group, Inc. v. Anthropic P.B.C. (M.D. Tenn.) – Musicians (Input Lawsuits)
- New York Times v. Microsoft (S.D.N.Y.) – News Publishers (Output Lawsuit)
- Getty Images v. Stability AI (D. Del.) – Photographers (Input Lawsuits)
- Doe v. Github, Inc. (N.D. Cal.) – Software Coders (Input Lawsuits)
It’s important to note that these lawsuits, which involve claims of copyright infringement related to using copyrighted works to train A.I. systems (Input Lawsuits) and the unauthorized use of AI-generated outputs (Output Lawsuits), are all still being litigated. This ongoing legal battle underscores the complexity and duration of the legal process in these cases.
Now, let’s delve into the significant concerns A.I. outputs pose for copyright law. Referencing the “Monkey Selfie” case, the critical issue is the distinction between human and machine authorship. For copyright to apply, a human must participate in the traditional elements of authorship. This requirement needs to be clearly defined and is subject to case-by-case interpretation.
Case Study: “A Recent Entrance to Paradise”
This case is one of the first to involve AI-generated work submitted to the Copyright Office. The work, titled “A Recent Entrance to Paradise,” was created autonomously by an A.I.AI called Creativity Machine, owned by Richard Taylor. Taylor submitted the job, claiming that while the A.I.AI created it, he should own the copyright as the A.I.AI’s owner.
The Copyright Office rejected the claim, stating that copyright requires human creativity, not machine creativity. Taylor could challenge this decision in court, but the court upheld the Copyright Office’s stance, confirming that a human must engage in intellectual, creative, and artistic labor to qualify for copyright protection.
AI Output Case Studies: Zarya of the Dawn
Zarya of the Dawn is a graphic novel with a post-apocalyptic, zombie-themed storyline created by Kris Kashtanova. Kashtanova used the AI tool MidJourney to generate the images for the book while she wrote the text and storyline herself. She made minor tweaks to some pictures before submitting the work to the Copyright Office, claiming full authorship.
Initially, Kashtanova received a copyright registration for the work. However, after publicizing her achievement on social media, the Copyright Office discovered her use of AI and decided to reevaluate the registration. They inquired into the details of her creative process, including the prompts she used and how MidJourney operates.
Upon review, the Copyright Office determined that the images were randomly generated based on her prompts, needing more human creativity. Consequently, they revised their decision, granting her copyright only for the text she wrote, not the images. Although they acknowledged her minor tweaks, these were deemed insufficient to meet the threshold for human creativity.
Kashtanova could challenge this decision in court, arguing the significance of her creative choices. However, she has not pursued legal action yet, so the case remains unresolved. This case, with its intricate interplay of AI, creativity, and copyright law, presents a fascinating intellectual puzzle for all involved.
SURYAST by Ankit Sahni using RAGHAV
- The artist took this photo actually clicked the shutter button and took a photo of the sunset and he stuck that into an AI machine he also plopped in an image of Van Gogh’s painting “Starry Night.”
- He told the machine take my photograph and make it in the style of “Starry Night” and then he told it how much transfer value he wanted it to to put into that process.
-
Then he submitted the output to the copyright office.
-
The copyright office says okay tell me what you did? You only really put in a transfer value and that’s not creative enough.
-
You might have copyright in your underlying work in the photograph.
-
But the output? You didn’t do anything for a color selection. You didn’t choose placement for the swirling pattern.
-
So this gave a little bit of guidance of what do you mean by traditional elements of human authorship?
-
Are you you are talking about color? You’re talking about placement? You’re talking about swirling patterns and texturizing?
-
So we are getting a little bit more verbiage, a little bit more guidance case by case, item by item, of what does the copyright office mean by human creativity?
-
This case was ultimately denied.
-
Again, we all have to wait and see if this case gets challenged in court.
AI Outputs – Other Jurisdictions
China – Mixed Protections and Disparities
Li v. Liu (Beijing International Court):
- Case: An artist used Stable Diffusion to create an image.
- Outcome: Copyright was granted to the artist using the platform.
Feilin v. Baidu (Beijing International Court):
- Case: Involved an AI-generated report containing tables and graphs.
- Outcome: No copyright was granted due to the lack of originality in the AI-generated content.
Tencent v. Yingxun (Shenzhen District Court):
- Case: Concerned about an AI-generated stock report.
- Outcome: Copyright was granted to the publisher based on the selection and arrangement of information.
These cases illustrate the varied approaches to copyright protection for AI-generated works in China, reflecting a complex and evolving legal landscape.
Consensus worldwide so far is that if you have enough prompts that are specific to color, placement, arrangement, environment, etc. MAYBE you can get a copyright but still no one can agree as to what the creative parameters are just yet.
AI Outputs – Deep Fakes
Digital Replicas
Recently, sexually explicit photographs of Taylor Swift, which were deep fakes, went viral on Twitter/X. In response, Congress is considering various bills imposing criminal and civil penalties for the unauthorized production of deep fake content.
Swift’s fans flooded social media with positive images of her in a show of support, using the hashtag #ProtectTaylorSwift
Name, Image, Likeness (NIL) Rights
Commonly referred to as NIL Rights, Rights of Publicity, or Rights of Privacy, these protections are sometimes described as a ‘haystack in a hurricane’ due to their complexity and variability. It’s important to note that these laws are not static but have evolved from torts and common law theories related to personal autonomy and the right to commercial use. Currently, no federal law governs NIL rights, but 34 states have enacted their own laws.
Critical Elements of NIL Laws
Most NIL laws encompass the following elements:
- Use of Someone’s Name, Image, Likeness, or Signature: The unauthorized use of an individual’s identity.
- Without Their Consent: The individual did not permit their identity use.
- For Some Benefit: The use is for commercial gain, such as making money or driving internet traffic.
Voice Misappropriation
OpenAI recently held a live demonstration of a new ChatGPT personal assistant voice, dubbed “Sky.” Observers noted a striking similarity between Sky’s voice and Scarlett Johansson’s in the 2013 film “Her.” OpenAI executives, however, denied any connection to Johansson.
A country singer discovered that someone had taken her published music, trained an AI on it, and used the AI to create a new song that she had never sung. This unauthorized song was then published online. In response, she made her own AI platform to license her voice.
Through her AI platform, the country singer has taken control over the usage of her voice. Users who wish to write music and create demos can now legally use their voice by paying a subscription fee. This not only generates revenue for her but also effectively addresses the issue of unauthorized use and potential legal actions.
Terms of Service/Terms of Use
Platforms increasingly incorporate specific clauses into their terms of service and use. For example, ChatGPT’s terms state that any potential rights in the AI-generated outputs belong to the user. However, all liability for using those outputs also falls on the user. If a third party claims the outputs infringe on their content and sues the platform, the user must indemnify it and cover its legal fees.
Additionally, some platforms require users to disclose when AI has been used to create content shared online. They often provide model language for these disclosures.
Meta, for instance, uses content from Facebook and Instagram to train its AI. However, Meta has paused these EU plans, which have stricter consumer protection and data privacy laws. In contrast, the U.S. does not have equivalent laws, allowing broader use without explicit permission.
There may also be terms restricting the commercial use of AI-generated content. It’s your responsibility to check for these clauses. Users should be proactive in differentiating between commercial and non-commercial use to ensure compliance.
Key Takeaways
- Refrain from assuming anything on the internet is free to use.
- Don’t copy-paste; use a hyperlink instead.
- Use proper copyright notices.
- Read the terms of service.
- Register your work promptly.
- Think in terms of risk avoidance.
- Stay informed.
- Get involved in the legislative process.
And…
Document Your Prompts
This is arguably the most important takeaway from this discussion. Especially for us artists working in entertainment, film, video games or commercials. In my opinion here’s why:
- AI Technology is Here to Stay: As a creative professional, these tools will soon become part of your daily workflow.
- Crux of Copyright Law: Prompts will be essential in determining if you have input enough creativity to make something unique.
- Creative Process: Document why you choose specific prompts (e.g., render type, camera choice) and any post-generation tweaks.
- Legal Protection: Detailed documentation will demonstrate your human creative contributions if you need to explain your process to the copyright office or a court.
- Risk Avoidance: The current legal landscape is uncertain, so use disclaimers and contract language to mitigate risks.
By thinking strategically and documenting your creative process, you can better protect your work and navigate the evolving field of AI-generated content.
I had hoped this talk would provide reasons why our roles as original creators wouldn’t shift to mere data entry. However, like the singer Grimes, we must accept this future and find ways to legally monetize it.