ChatGPT: Do AI-based Texts Violate Copyright?
Artificial intelligence that writes texts on its own: What sounds like science fiction is already reality and and has reached the general public with tools like ChatGTP. However, this world full of new possibilities in content creation also raises previously unknown questions for users in the area of copyright. Are works of artificial intelligence protected by copyright? What do you have to consider in the area of copyright when you publish AI-based content yourself in order to avoid a liability situation? What rights are you entitled to as the author? We answer these questions and more in the article.
AI-based Content - More than a Pipe Dream
The topic of artificial intelligence and the resulting issues have long been something that was only discussed in expert circles. However, the AI-based chatbot tool ChatGPT has changed things and the topic has moved into the focus of public perception. The idea of having content created by an artificial intelligence is not entirely new - various companies such as Neuroflash or image generators such as Jasper or Runway were experimenting more or less successfully with different text generators for years.
These technologies definitely offer great potential, but also carry the risk of legal conflicts and the resulting liability. Firstly, this applies when you use the AI-generated results of a platform, also for the training data used to feed the AI for learning purposes. But before we turn to these points, let's focus on a few basic explanations about ChatGPT.
What is ChatGPT?
Developed by the US company OpenAI, ChatGPT (GPT = Generative Pre-trained Transformer) has been shaking up the ranks of AI-based tools since its release in November 2022. The chatbot works as a text-based dialogue system in the form of a user interface and is based on machine learning. Within the framework of the Reinforcement Learning from Human Feedback (RLHF) method, the tool was repeatedly audited, with the output of the chatbot being evaluated by humans. ChatGPT takes this feedback into account for future work.
ChatGPT: This Is How It Works
In order to keep learning, an artificial intelligence needs so-called training data in addition to human feedback. In the case of ChatGPT, this data primarily consisted of texts created by humans, for example from social media, books, newspapers or voice recordings. In this way, the chatbot was supposed to learn to imitate human responses as authentically as possible.
As a user, getting in touch with ChatGPT is very simple: the AI can be reached via its own login page. After entering some data such as your email address and telephone number, you can contact the chatbot directly and request the creation of various types of text - from technical articles to narrative writing to computer code.
Copyright Protection For AI-based Works?
In principle, anyone who has created a work holds the rights to it in the first place. But does that also apply to an artificial intelligence? From a legal point of view, the answer is: Rather not. Because if you let an AI create content, it is actually extremely questionable who ultimately owns the rights to this content. Considering only personal intellectual works as works in the legal sense supports the thesis that AI-driven content is lacking human creativity. This means that content created by ChatGPT is not protected by copyright in the first place. Nor are you personally entitled to any copyright after you have submitted a query to ChatGPT. This is because entering a request into a text window does not count as a creative act.
Copyright protection not only exists for images and text, but also for musical works. An IT service provider who was supposed to provide the background music for a fashion event found out how strict these regulations are.
Other interpretations, however, state that copyright protection may exist if the human influence in the course of creating something is so great that the final result can be attributed to it rather than to the chatbot. But, such a subordinate role is not the case with ChatGPT and it can be assumed that texts generated via this tool do not enjoy any protection. Yet can you, a third party, then also use these works as you wish?
Free Use Of AI-based Works?
It seems so simple: you need a text, have it conveniently created by ChatGPT and then use it as you wish, for example for your website. But it is important to consider how the chatbot generates the answers. This is because the AI does not build texts completely generically, but creates content based on training data. In the case of ChatGPT, this is usually human-made documents and recordings - and these enjoy copyright protection.
If ChatGPT uses this data to generate text it is a duplication. In practice this means either the use of existing text structures or in modified text extracts at the end of the process. If you now access this content and, for example, upload it to your website, you make the content accessible to others. While an AI like ChatGPT is unlikely to be legally classified as the original author of content, the chatbot's output is based on human-generated text, as mentioned above. Ostensibly, these texts are public domain or licensed, but it is difficult for a user to verify this. Using the generated texts without verifying them creates a legal minefield, and infringements of copyright through the unlawful use of other people's content are inevitable.
If the worst comes to the worst and you receive a written warning, you can usually prevent something worse by taking the right action. Our article Have You Received a Written Warning? This Is How You Should Respond reveals what is important.
Our recommendation: Use ChatGPT to make your work easier, but do not base all your content creation on this tool. In addition, you should regularly check the chatbot's output for originality to avoid the risk of plagiarism. This can be done, for example, with free tools such as PlagAware.
Your Content As AI Training Data: How To Protect Yourself
Of course, your own text could be used as training data for ChatGPT, appear in the chatbot's output and then be misused by others, in addition to the careless use of AI-generated content. Text and data mining has been further liberalised over the years. This is the automated analysis of digital works to extract patterns, trends and correlations - an important process for 'feeding' AI with training data. In many cases, the automatic collection of data through websites for commercial purposes is also allowed now.
Not only media professionals are exposed to the risk of infringement. An IT service provider also had to experience how quickly a written warning can hit him: Trademark Infringement: An IT Service Provider Forgets to Do his Trademark Research.
For example, if you operate a website from Estonia, you can at least protect yourself to some extent against this in the form of an opt-out in machine-readable form. The best way to do this, is by using a robots.txt file. With this file, you tell crawlers (programmes that independently search the internet for content) which parts of your website they are allowed to read and which they are not. However, a robots.txt file does not offer one hundred percent of protection against access. If you want to be on the safe side, you should provide your web server with password protection.
Are Your Texts Used Illegally?
Prevention is great, but what do you do if your text has already found its way into the output of ChatGPT and is now being used illegally? You can find a whole range of free tools online to help you determine whether your texts have been illegally taken. Copyscape is particularly widespread. There you can enter the URL of your website and the tool will determine whether your content appears elsewhere on the net.
If you now discover content theft, you have various options at your disposal. Your first step should be to save the evidence in the form of screenshots with dates and times. If you want the illegally used content removed, be careful: The offender may have accessed ChatGPT-generated text in good faith and be unaware of any wrongdoing. Your first step should be to simply ask for your content to be deleted. If that doesn't work, you may want to consider hiring a lawyer who can sue the infringer to stop and claim damages on your behalf.
AI-Generated Content: Lots Of Potential, Lots Of Risks
ChatGPT offers many industries from Estonia, from IT to journalism, the opportunity to make their work enormously easier. But until urgent legal questions are finally settled, there are too many uncertainties that render the unconditional use of the tool a high legal risk - too great are the risks of copyright infringement and the resulting costly written warnings and claims for compensation.
Therefore, you should refer to the existing copyright law when making your decisions and exercise great caution when using AI-based content.
With Or Without ChatGPT: Protect Yourself Against Written Warnings!
Whether due to the use of AI-based content or other negligence: If third-party content is illegally placed on your Estonian website, the risk of a written warning including costly claims for compensation is significant. It is good if you do not have to face this dispute alone and bear the costs yourself. With Professional Indemnity Insurance through exali, you are covered against the consequences of legal infringements - even if they are caused by the use of AI tools.
In the case of unclear liability, the insurer checks the claims made against you, fends off unjustified demands and bears the costs for justified claims Put together your insurance coverage right away - our customer service team will be happy to assist you from Monday to Friday from 9:00 a.m. to 6:00 p.m. (CET) on +49 (0) 821 80 99 46-0.
Vivien Gebhardt is an online editor at exali. She creates content on topics that are of interest to self-employed people, freelancers and entrepreneurs. Her specialties are risks in e-commerce, legal topics and claims that have happened to exali insured freelancers.
She has been a freelance copywriter herself since 2021 and therefore knows from experience what the target group is concerned about.