Large Language Models, Artificial Intelligence and the Future of Law

How do we conceptualise IP in an AI age?

Who owns copyright to such AI generated content?

• The Developer?

• The Inspirer?

• Joint-holding of training data holders?

• The End user?

What does the law say?

United States

"The Office explained that human authorship is required for copyright protection, that it will refuse to register works solely generated by AI, and that any material which is not the product of human authorship must be disclaimed on a registration application"

United Kingdom

In the case of a literary, dramatic, musical or artistic work which is computer-generated, the author shall be taken to be the person by whom the arrangements necessary for the creation of the work are undertaken. 

India

(d) “author” means, — (i) in relation to a literary or dramatic work, the author of the work; (ii) in relation to a musical work, the composer; (iii) in relation to an artistic work other than a photograph, the artist; (iv) in relation to a photograph, the person taking the photograph; 2 [(v) in relation to a cinematograph film or sound recording, the producer; and (vi) in relation to any literary, dramatic, musical or artistic work which is computer-generated, the person who causes the work to be created;] 

What do AI companies say?

Open AI

You may provide input to the Services (“Input”), and receive output from the Services based on the Input (“Output”). Input and Output are collectively “Content.” You are responsible for Content, including ensuring that it does not violate any applicable law or these Terms. You represent and warrant that you have all rights, licenses, and permissions needed to provide Input to our Services.

Ownership of content. As between you and OpenAI, and to the extent permitted by applicable law, you (a) retain your ownership rights in Input and (b) own the Output. We hereby assign to you all our right, title, and interest, if any, in and to Output. 

Similarity of content. Due to the nature of our Services and artificial intelligence generally, output may not be unique and other users may receive similar output from our Services. Our assignment above does not extend to other users’ output or any Third Party Output.

Google

Google on October 12 said that it would fight any copyright lawsuits filed against customers that use its generative AI services. “Put simply: if customers are challenged on copyright grounds, we will assume responsibility for the potential legal risks involved,” the company said.

Microsoft

Microsoft announced its new ‘Copilot Copyright Commitment’ saying that the company will offer protection from Copyright claims to its customers using its AI-powered Copilot services. On September 7, the company stated in a blogpost that it will “assume responsibility for the potential legal risks involved” if customers are challenged on Copyright grounds for using the output generated by Microsoft’s generative AI services. 

Does training AI models violate copyright of authors?

A litany of lawsuits

1. November 2022: A class action lawsuit against Microsoft, GitHub, and OpenAI alleged that GitHub Copilot, an AI-powered code editing tool trained on public GitHub repositories, violated the copyright of the repositories' authors by generating source code that matched its training data verbatim without attribution. 

2. January 2023: Artists Sarah Andersen, Kelly McKernan, and Karla Ortiz filed a class action lawsuit against Stability AI, Midjourney, and DeviantArt, claiming copyright infringement for training AI tools on five billion images scraped from the web without the artists' consent. Criticisms of the lawsuit noted technical inaccuracies about AI capabilities. 

3. January 2023: Stability AI faced a lawsuit in London from Getty Images for using its images in their training data without purchasing a license. 

4. February 2023: Getty Images filed another lawsuit against Stability AI in a U.S. district court in Delaware, alleging copyright and trademark infringement for using Getty's images in training data and generating images with Getty's watermark. 

5. July 2023: Authors Paul Tremblay and Mona Awad filed lawsuits against OpenAI, alleging unauthorized use of their copyrighted books by ChatGPT. Further lawsuits were filed by Sarah Silverman, Christopher Golden, and Richard Kadrey against Meta and OpenAI over similar issues. 

6. December 2023: The New York Times sued Microsoft and OpenAI, claiming that their AI engines were trained on articles from the Times, which they considered an infringement of their copyright. =

7. April 2024: Eight U.S. national newspapers owned by Tribune Publishing sued Microsoft and OpenAI for copyright infringement related to using their articles in training data and for outputs creating false statements attributed to these newspapers. 

• In India, Digital News Publishers Association (DNPA), an umbrella organization working in the interest of news publishers, sent a letter to the Central Government seeking copyright protection rules against generative Artificial Intelligence (AI) models.

What does the law say?

United States

In the US, the principle of "fair use" is often cited as a defense in copyright infringement lawsuits, including those involving AI. Fair use is determined on a case-by-case basis, considering factors such as:

• Purpose and Character of the Use: Transformative uses (which add new expression or meaning) are more likely to be considered fair use. This can include using copyrighted materials to train AI models that generate new, original content.

• Nature of the Copyrighted Work: Uses of factual or less creative works are more likely to be seen as fair use compared to highly creative works.

• Amount and Substantiality: Using smaller amounts of the original work or parts that are not considered the "heart" of the work may favor fair use.

• Effect on the Market: If the use does not harm the market for the original work or its potential markets, it may be considered fair use.

AI companies might argue that their use of copyrighted materials to train algorithms is transformative and does not negatively impact the market for the original works. However, the specific circumstances of how the data is used and the nature of the output generated are crucial. See Authors Guild, Inc. v. Google, Inc.

United Kingdom

In the UK, copyright law includes certain exceptions that might be relevant for AI, such as:

• Incidental Inclusion: If a copyrighted work is incidentally included in another work, this may not constitute infringement.

• Research and Private Study: Use for non-commercial research and private study may be allowed.

• Text and Data Mining (TDM): There is an exception for copying for the purpose of computational analysis for non-commercial research, provided there is lawful access to the work.

AI companies could potentially leverage the TDM exception, arguing that the use of copyrighted texts and images to train models falls under computational analysis. However, this would depend on the nature of the AI application and whether it is deemed commercial.

India

• Fair Dealing: Similar to fair use, fair dealing in India allows for the use of copyrighted material without permission for criticism, review, reporting, research, or private study. However, this is more limited compared to the US fair use.

• Use for Teaching: Use in instructional settings might not constitute an infringement if it is justified by the educational purpose.

• Incidental or Accidental Inclusion: If the inclusion of a copyrighted work is incidental or accidental, it might not constitute infringement.

AI companies in India might use the fair dealing exception for training AI models, especially in educational or research contexts. However, the commercial nature of many AI applications complicates this defence.

Do we need a special law for AI and Intellectual Property?

• In the US, special legislation such as the Generative AI Copyright Disclosure Act has been proposed. The copyright office has also started consultations of the relationship between AI and copyright. 

• In the UK, the IP office is trying to implement an AI/copyright code of practice after consultation with all stakeholders. 

• In India, the 161st Parliamentary Standing Committee Report also found that the Copyright Act of 1957 isn't well-equipped to allow AI to own and create authorship. However, the Minister of State stated that while Artificial Intelligence (AI) and related innovations is an evolving stream of technology the current legal framework under the Patent and Copyright Act is well-equipped to protect Artificial Intelligence generated works and related innovations.