Generative AI and Copyright Infringement: What Brands Need to Know
Generative AI has revolutionized content creation for businesses. But due to concerns over copyright infringement, these tools are now the subject of numerous legal battles.
Generative AI: Opportunity or Risk?
Generative AI platforms empower businesses with the ability to create text, images, and videos with just a few prompts. These AI-driven tools produce content almost instantly, providing companies with creative material that’s ready for use within minutes.
These solutions help companies save time and manpower hours, as businesses can now simply type in commands describing the content they want – no creative team required. With AI, it’s possible that one person can effectively do the job of many people and rapidly scale content production.
But the use of these tools is accompanied by a host of complex issues. Because laws around copyright and Generative AI are constantly evolving, it’s critical brands stay informed of the latest developments in order to protect their intellectual property, as well as avoid potential legal issues.
Unintentional copyright infringement is a serious risk that comes with using automated content creation tools. AI tools generate large volumes of content, such as text, images, and music, but some of these results may unintentionally replicate or closely resemble copyrighted material.
AI-driven deepfakes or synthetic media are also a challenge. Generative AI can create convincing-looking content that misuses an established brand’s identity, logo, or copyrighted content without permission. Clearly, that can lead to unauthorized use and legal issues. These no-copyright AI images used by brands may, in fact, be violating local copyright laws.
Theoretically, a copyright holder could seek damages against a company using AI-generated material that mimics the original content. With the rapid adoption of AI, we are seeing more of this situation. So, increasingly, copyright holders are realizing that their content has been used without permission. This has of course, resulted in an uptick in legal action stemming from the use of AI. But to properly understand how copyrighted material has been used differently than in the past and how AI has become the subject of multiple lawsuits, we need to take a brief look at how AI functions.
Understanding the Role of Generative AI
Like nearly all AI tools, Generative AI tools are based on algorithms leveraging AI models that are trained on vast datasets. These datasets, which often include copyrighted works, allow the models to learn styles, language, and patterns, enabling them to generate diverse content across multiple mediums. This means there can be the use of protected content without the proper licensing.
Popular generative AI tools include ChatGPT, DALL-E, Midjourny, and Stable Diffusion. ChatGPT is a text-based tool that creates written content - including code - based on user prompts. DALL-E is a generative AI model developed by OpenAI, the creators of ChatGPT, which produces detailed images based on user prompts.
Similar to DALL-E, Stable Diffusion and Midjorney are tools that create original images based on text prompts. In addition to this, they also produce images with image prompts in order to reproduce certain visual elements. However, unlike DALL-E and ChatGPT, Stable Diffusion is specifically an open-source tool, meaning that it allows developers and researchers to modify its models for various uses. Stable diffusion has become the popular AI model for multiple programs and, in some cases, allows users to train the AI to reproduce a specific artist’s style.
The Copyright Conundrum
The recent surge of AI copyright cases demonstrates the complicated legal issues posed by the use of Generative AI, especially when it comes to infringement. While some AI companies have argued that training their models on copyrighted works does not constitute infringement because it’s considered fair use, that position has not been completely accepted by courts. Precedent-setting cases on AI and copyright infringement are still being litigated.
Copyright holders and legal practitioners have said that while the infringement may be unintentional, the creation of content that closely mimics or is virtually identical to protected intellectual property might in fact constitute infringement.
Generative AI Copyright Infringement: Real-World Examples
The Walt Disney Company’s concerns over Microsoft's Bing AI Imaging Tool highlight the risk of AI-generated content potentially infringing on trademarks. The Bing AI tool, powered by DALL-E 3, allowed users to create images in a "Pixar" style. Disney, which owns Pixar, expressed its disapproval of the tool because it included the iconic Disney-Pixar logo within the original images it created.
This is just one example of how AI-generated content produced by one company could inadvertently infringe on the trademarks of another company, even without intending to do so. Microsoft settled the issue by tweaking its models, resulting in a garbled version of the Pixar logo being produced.
AI researchers ran an experiment in which they discovered popular text-to-image AI tools often replicated nearly exact copies of copyrighted images, such as famous movie characters. The test was replicated by multiple experts, who also found that their prompts resulted in images virtually identical to copyrighted works.
In case this isn’t enough of a red flag, Patronus AI – an AI model evaluation company –released research showing OpenAI’s GPT-4 regularly produces copyrighted content because it responded to an average 44% of prompts asking to reproduce copyrighted text. While various lawsuits have so far been from authors who oppose training generative AI models using their copyrighted works, we see it’s possible for occurrences of the actual produced content to become the subject of future lawsuits.
Legal and Ethical Implications of Generative AI
While the legal landscape regarding AI-related intellectual property infringement is still evolving, companies whose work is at risk of being copied are becoming more vigilant. The recent lawsuit filed by Getty Images against Stability AI for using their images without licensing or permissions underscores the potential for AI tools to use copyrighted material without authorization.
The impact of AI-generated content on original content creators raises ethical concerns. Because AI tools often replicate or adapt protected works without proper credit, this could lead to a loss of the creator’s intellectual property rights and/or income. This means brands should be aware of the risks that AI-generated content poses to their intellectual property. While the laws are still vague and evolving, it currently puts the onus on creators or brands to make sure they are protecting the proper use of their content.
Guardrails and Solutions for Generative AI Copyright Infringement
There are a number of approaches currently being considered to solve issues around AI and copyrighting. Establishing better legal frameworks that clarify the rights and responsibilities of AI developers, users, and content creators is being pursued as one solution. Additionally, watermarking (text or image to show ownership and prevent unauthorized use), embedding metadata and other technical solutions may help determine whether AI-generated content is based on copyrighted material.
Much work remains to be done on the issue. The “Snoopy” problem refers to a specific Generative AI challenge, illustrating just how challenging it is to protect copyrighted material. The more distinctive a copyrighted image, like the cartoon character Snoopy, the higher the likelihood that a Generative AI tool will replicate that content.
Policy and Regulation Impacting Generative AI
Both the EU and the US are currently working on establishing legal frameworks to govern the use of Generative AI, including copyright infringement issues.
The EU’s AI Act requires that AI models adhere to strict requirements, including transparency, accountability, and minimizing copyright risks. The act mandates that Generative AI systems comply with these rules, which are pro-copyright protection.
In October of 2023, the US released a long-awaited Executive Order on the Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence (EO). This EO attempts to address copyright and intellectual property issues while providing the starting point for further legislation.
After the 2023 EO in the US, individual states and federal agencies have approached the challenge via different frameworks. In March 2024, the U.S. Copyright Office launched an initiative specifically focused on AI copyright infringement. The project also aims to determine if AI images can be copyrighted.
Brands Should Guard Their Intellectual Property from Generative AI
The benefits of Generative AI are clear. They produce compelling content that companies need to stay relevant in today’s digital landscape, lowering the barrier to entry for content creation and helping businesses save time and money when producing original materials.
But the potential risks of Generative AI, particularly around copyright infringement, are significant. As the use of generative AI increases, so does the risk that AI-created content contains copyrighted material. So, while the legal framework and preventative measures haven’t yet caught up, brands with intellectual property should be aware that their copyrighted work could quickly surface elsewhere. And, of course, the more users of these generative AI tools, the more possibility this content will be created.
BrandShield’s robust brand protection platform helps businesses safeguard their intellectual property, including in cases of copyright infringement stemming from Generative AI. To learn more about how we can empower you with the tools you need to ensure that your brand’s copyright and intellectual property are safeguarded from AI-generated content, talk to us today.
