Generative Artificial Intelligence and Current Issues in Copyright – Publications

Recent popular applications powered by generative artificial intelligence, such as ChatGPT, create important copyright issues for individuals and businesses regarding content creation, including commercial use, content publication, copyright infringement, and content limitation. enforcement

After years of development and hope, overnight artificial intelligence was struck.

Recently, several applications powered by Artificial Intelligence (AI) have gained popularity among content creators around the world who are learning to harness the power of generative AI. AI powered models like ChatGPT, Dall-E and Midjourney are in the news every day and many people have become dependent on their capabilities, and are amazed. These applications are revolutionizing the ease with which many types of content can be quickly created, including website copy, marketing campaigns, social media posts, lists of all kinds, poetry and software code.

One of the most popular AI tools, ChatGPT is a chatbot created by OpenAI in San Francisco and can generate human-like text in response based on deep learning techniques, which involves training an AI model on a large dataset. Text to predict the patterns and relationships of words and phrases. In fact, the opening paragraph above begins with a strong outline of ChatGPT, where we push to describe what ChatGPT is and can do. After a little tweaking, we got what we were looking for.

ChatGPT’s free language model GPT-3 was released in June 2020. However, OpenAI has released its fourth iteration of the software, GPT-4. [1] OpenAI’s website [2] He explains that GPT-4 is “more creative and collaborative” and even allows AI to “create, edit and collaborate with users on creative and technical writing tasks, such as writing songs, writing screenplays or learning the user’s writing style”. of Abajornal He recently published an article titled “Latest ChatGPT aces bar test near 90th percentile”. [3] While ChatGPT was limited to generating text responses, GPT-4 now allows users to provide images as input and can even generate comments or code based on the provided image. Other OpenAI applications such as Dall-E and Midjourney provide text-to-image models.

The fine print

As with most applications, the terms of use are the starting point for understanding the scope of rights granted to the user. The Terms of Use of ChatGPT refer to the following:

• Content Ownership: “Content” is defined in OpenAI’s Terms of Use (Terms) for ChatGPT as any input or output generated by the Service. In relation to ownership, the regulations state:[a]As between the parties and to the fullest extent permitted by applicable law, you own all Submissions, and subject to your compliance with these Terms, Open AI grants you right, title, and all rights to and to the results thereof. OpenAI retains limited rights to use Content to provide and maintain the Service, to comply with applicable law, and to enforce OpenAI’s guidelines. [4]
• Similar results: The terms warn that output may not always be exclusive or exclusive to one user:
  • “Due to the nature of machine learning, output may not be unique to users and the Services may produce the same or similar results to OpenAI or a third party. For example, such as, ‘What color is the sky?’ You can provide input to the model and receive results such as “The sky is blue.” Other users may ask similar questions and receive similar responses. Responses asked and generated by other users are not considered your content.
• Confidentiality: Content submissions to ChatGPT are disclosures, and may be used to train the models as described in the Terms’ Proprietary Provisions, and certain amounts of Content may be part of other ChatGPT responses. The lack of privacy associated with using ChatGPT is the main reason why companies should be cautious about using it by employees.
• Publication Requirements: OpenAI’s Publishing Policy [5] Responds directly to requests for further publication of results: A company wishing to publish “is permitted to use written content created in part by ChatGPT (e.g., a book, a collection of short stories) under the following conditions: (1) the published content belongs to the company, (2) the role of AI in creating the content It is clearly stated that (3) Content topics do not violate OpenAI’s Content Policy [6] or Terms of Use, (4) [OpenAI] Politely ask that you refrain from sharing results that might offend others.

OpenAI also explains “the role of AI [must be] It is clearly explained in a way that no reader can miss and a common reader can easily understand. The rules provide a sample statement:

For example, one should list the relative roles of drafting, editing, etc. in the foreword or introduction (or similar place). He is the person who should take ultimate responsibility for the content that is published.

ChatGPT provides sample language to describe a typical creative process:

The author created this article in part with GPT-3, OpenAI’s large-scale language-generation model. When generating draft language, the author reviews, edits, and revises the language to their own needs and assumes ultimate responsibility for the content of this publication. [7]

Possible risks

Although OpenAI is unlikely to object to the company’s use of content generated by ChatGPT (unless the user violates its terms), that does not mean that use of such content is without the risk of copyright-related challenges.

As long as AI models are relied upon to generate content, liability for copyright infringement still exists. Both artists and businesses have filed claims in federal court over the use of copyrighted images to train AI models and allegedly infringe their copyrights.

For example, in a class action involving “at least thousands” of other creators, Midjourney, Stability AI and DeviantArt are accused of copyright infringement, infringement of publicity rights and unfair competition. [8] Getty Images has filed a lawsuit against Stability AI in the United States and separately in the United Kingdom, alleging they used more than 12 million images from the Getty database for machine learning purposes.[9] AI software code generation tools are similarly tested. [10]

In these cases, defendants claim that their use of copyrighted material in this manner constitutes fair use. These cases are still in their early stages, but whether importing copyrighted material for machine learning purposes is infringement or fair use may eventually need to be decided by the US Supreme Court. Accordingly, OpenAI may be prosecuted based on the use/access of third-party content to train ChatGPIT, and depending on the outcome of any circumstances, certain users may be prosecuted if there is content generated by ChatGPIT. (and subsequently used by a particular user) is very similar to third-party content.

Waiting for the results of the creative AI

The recent US Copyright Office decision regarding the copyright of an AI-generated graphic novel titled “Zarya of the Dawn” is instructive as to how much and to what extent companies can protect content created using ChatGPT and other similar AI tools.

“Zaria of the Dawn” is a comic book with original text written by a human but created by Midjourney, a fast-paced AI program to generate images. The resulting images were then selected, edited, and partially edited by a human author. The Copyright Office initially allowed the work to be registered, but when it found out that AI-generated images were used – it revoked the copyright and limited the scope of the copyright to cover only the original text and the final product. Protection for individual AI-generated images.

The Copyright Office requires human authorship for creative works to be copyrighted. When AI is used in cases like “Zarya of the Dawn,” the Copyright Office looks closely at human and non-human contributions. While the author argued that the creative process involved a person guiding the AI ​​through a series of hundreds of intermediate images until each image achieved the desired result, the Copyright Office explained that a person submitting a text message would lead the AI. A machine “does not actually create the images it creates and is not the ‘mastermind’ behind them.” in other words, “[t]The information in the prompt may have an ‘influence’ on the resulting image, but the prompt text does not specify a specific effect.

The Copyright Office decision further clarified the issue of human control over generative AI compared to the comparative control a photographer has over a camera. According to the office, unlike the photographer, AI models do not have the same human control over the final images produced, such as framing, lighting, exposure, depth, etc. The author of “Zaria of the Dawn” does not control how the subject matter or presentation of the subject matter appears, and cannot control what the AI ​​creates, as discussed in the Copyright Office’s decision.

The copyright office says, “While additional queries applied to one of these initial images may affect subsequent images, the process is not under the user’s control because it is impossible to predict in advance what Midjourney will create.” Moreover, the Copyright Office has stated that the author’s editing of images is “too small and incomprehensible” to satisfy the creativity necessary for copyright protection.

The Copyright Office recently released a policy statement that provides additional guidance on whether and to what extent artistic works created with the help of generative AI are eligible for copyright protection. [11] In direct response to the Office’s ruling on “Zaria of the Dawn,” going forward, “the Office will examine whether AI contributions are ‘mechanical reproductions’ or an author’s ‘own original mental concept.’ [the author] It gave visible form.’” Specifically, “AI technology determines the descriptive elements of the output, not the output of a human author.

The Office’s guidance specifically states that “[b]Based on the Office’s understanding of the currently available generative AI technologies” users currently Lack of necessary creative control. It has yet to be determined whether GPT-4 or other AI tools going forward will give users enough creative control over AI output to be considered registered for that output in the Copyright Office. However, unlike traditional issues related to the co-creation of works, with AI output, there will be a perfect written record of the creative process. Will AI service providers one day be subpoenaed for chat records to prove or disprove the adequacy of human authorship?

Summary

With their ease of use and, in many cases, impressive results, ChatGPT and similar AI services are likely here to stay. As generative AI platforms expand and improve, it will be important for businesses to actively think about how these services should and should not be used in the workplace. Morgan Lewis is ready to help readers develop appropriate AI content usage policies for their businesses.