Authorship and Journalism in the Era of Generative AI

[1] 

AI services are increasingly the subject of lawsuits by creators seeking to protect their creative works.[2] A consolidated class of fiction authors, non-fiction authors, and The Authors Guild filed a complaint against OpenAI and Microsoft in the United States District Court for the Southern District of New York on February 5, 2024.[3] The class includes authors such as George R.R. Martin, Elin Hilderbrand, and Jodi Picoult, among others.[4] The Authors Guild describes itself as “the oldest and largest professional organization for published writers,” with over 14,000 members.[5] Plaintiffs have alleged that Defendants’ generative AI programs “mimic[], summarize[], and paraphrase[]” the works of these authors, amounting to “systemic theft on a mass scale” of copyrighted materials in the training of OpenAI’s Large Language Models (LLMs), which power ChatGPT’s generative capabilities.[6] In one episode that the Plaintiffs called attention to in their complaint, a developer allegedly utilized ChatGPT to draft versions of two forthcoming novels by George R.R. Martin, “The Winds of Winter” and “A Dream of Spring” from his bestselling series, A Song of Ice and Fire.[7]

In another lawsuit filed in December of 2023 in the Southern District of New York, The New York Times has alleged that OpenAI and Microsoft’s generative AI programs “recite[] Times content verbatim, closely summarize[] it, and mimic[] its expressive style.”[8] The Times alleges this results from OpenAI’s copying and use of the Times’ works in LLM training.[9] The Times has also asserted that Microsoft’s Bing Chat, powered by ChatGPT-4, produced “hallucinations” that misattributed information to the Times that was never published, thus causing the Times commercial injury.[10] In one instance, the Times alleges Bing Chat created specific quotes attributed to Moira Forbes that never appeared in The Times article in question or anywhere else.[11]

Training of ChatGPT’s LLMs has expanded with each iteration; ChatGPT-3 hosted 175 billion parameters and was 100 times larger than GPT-2.[12] The most recent iteration is ChatGPT-4, released in March of 2023.[13] OpenAI has closely held the data, computing power, and training techniques of this most recent model.[14]

According to the Authors Guild, OpenAI has described the purpose of training their LLMs as to teach ChatGPT “how words fit together grammatically” and how words come together to “form higher-level ideas.”[15] In response to the class action case brought by the authors, OpenAI asserted that using the plaintiffs’ books to teach their LLMs intelligence and language would be “paradigmatic transformative fair use.”[16] As such, the court will likely be tasked with applying the “fair use doctrine” in which the court will weigh the “fair use” factors in determining whether OpenAI’s training should be afforded protection under 17 U.S.C. § 107.[17]

The Times is concerned that unrestrained generative AI will lead to news organizations having fewer journalists and fewer resources to develop quality journalism, creating a “risk that stories will go untold.”[18] The class-action of authors, on the other hand, fear that the trajectory of generative AI training absent proper deference to creators will lead to distortion of the market for professional writers’ work and a rise of derivative works.[19] Generally, courts have been deferential to tech companies in determining how copyright law applies to artificial intelligence.[20] Regardless, these cases have broad implications for the future of generative AI. As demonstrative of the existential implications for AI models, in The Times’ proceedings, it has requested the court to order the destruction of any GPT that has incorporated the Times’s works.[21]

References:

[1] Photograph of a Book, in Stockvault (last visited Mar. 31, 2024), https://www.stockvault.net/photo/192034/book.

[2] Blake Brittain, Authors Suing OpenAI Lose Bid to Halt Rival N.Y. Copyright Lawsuits, Reuters (Mar. 4, 2024) https://www.reuters.com/legal/litigation/authors-suing-openai-lose-bid-halt-rival-ny-copyright-lawsuits-2024-03-04/ (explaining how different groups of copyright owners have sued tech companies over the training of large language models).

[3] First Consolidated Class Action Complaint at ¶ 1-2, Authors Guild v. OpenAI, No. 1:23-cv-08292-SHS (S.D.N.Y. filed February 5, 2024) (providing identifying information for parties and nature of case).

[4] Id. at 2 (specifying members in class of authors).

[5] About the Guild, The Authors Guild (last visited Feb. 24, 2024), https://authorsguild.org/about/ (providing background on The Authors Guild).

[6] See First Consolidated Class Action Complaint, supra note 1, ¶ 2-5 (describing consolidated class’s copyright infringement allegations).

[7] Id. ¶ 279; see also Rebekah Valentine, Someone Used ChatGPT to Finish the Game of Thrones Book Series, IGN (July 19, 2023), https://www.ign.com/articles/someone-used-chatgpt-to-finish-the-game-of-thrones-book-series (explaining developer, Liam Swayne’s drafting process of books on ChatGPT).

[8] Complaint, ¶ 4, New York Times v. Microsoft, No. 1:23CV11195 (S.D.N.Y. filed December 27. 2023) (alleging nature of AI program output). ChatGPT-4 powers Copilot (formerly Bing Chat), Microsoft’s generative AI service. Heaven, infra note 11. OpenAI has integrated the ability to search the internet in ChatGPT utilizing Microsoft’s technology in “Browse with Bing.” Aaron Mok, ChatGPT is Getting an Upgrade that Will Make it More Up to Date, Business Insider (Nov. 6, 2023), https://www.businessinsider.com/open-ai-chatgpt-training-up-to-date-gpt4-turbo-2023-11#:~:text=ChatGPT%20users%20will%20soon%20have,information%20up%20to%20April%202023.  OpenAI was founded in 2015 as a non-profit and launched a for-profit enterprise in 2019. Charles Duhigg, The Inside Story of Microsoft’s Partnership with OpenAI, The New Yorker (Dec. 1, 2023), https://www.newyorker.com/magazine/2023/12/11/the-inside-story-of-microsofts-partnership-with-openai.  Since then, Microsoft has allegedly invested $13 billion from 2019-2023. See Id.

[9] Id. ¶ 4-5 (describing Times’s allegations regarding how OpenAI’s output reflects training of copyrighted material).

[10] Id. at ¶ 135-142; see also Eileen McDermott, New York Times Takes on OpenAI, Microsoft, IPWatchdog (January 2, 2024), https://ipwatchdog.com/2024/01/02/new-york-times-takes-openai-microsoft/id=171355/ (providing background on AI “hallucinations” and how they have appeared in BingChat).

[11] Complaint, supra note 6, ¶ 138 (alleging a specific instance where BingChat misattributed content to The New York Times).

[12] Will Douglas Heaven, GPT-4 is Bigger and Better than ChatGPT – But OpenAI Won’t Say Why, MIT Technology Review (Mar. 14, 2023), https://www.technologyreview.com/2023/03/14/1069823/gpt-4-is-bigger-and-better-chatgpt-openai/ (noting progression of training parameters from ChatGPT-2 to ChatGPT-3).

[13] Id. (explaining the release date of ChatGPT-4).

[14] See id. (discussing OpenAI’s secretive approach to GPT-4); See also Complaint, supra note 7, ¶ 77 (alleging a quantity of 1.76 trillion parameters in GPT-4). 

[15] First Consolidated Class Action Complaint, supra note 1, ¶ 5 (asserting OpenAI’s expressed purpose for training LLMs).

[16] OpenAI Defendants’ Answer to First Consolidated Class Action Complaint, at 3, Authors Guild v. OpenAI, No. 1:23-cv-08292-SHS (S.D.N.Y. filed Feb. 16, 2024) (asserting fair use defense in response to plaintiffs’ copyright claim).

[17] See Edward D. Lanquist & Jeremy D. Ray, Artificial Intelligence and Copyright Law: The NYT v. OpenAI – Fair Use Implications of Generative AI, Baker Donelson (Feb. 5, 2024), https://www.bakerdonelson.com/artificial-intelligence-and-copyright-law-the-nyt-v-openai-fair-use-implications-of-generative-ai (listing factors for fair use defense).

[18] Complaint, supra note 6, ¶ 48 (enumerating The Times’s concerns regarding journalism in face of generative AI).

[19] See First Consolidated Class Action Complaint, supra note 1, ¶ 3 (describing Authors Guild’s and plaintiff authors’ concerns regarding AI impact on professional writing).

[20] See Matt O’Brien, ChatGPT-Maker Braces for Fight with New York Times and Authors on “Fair Use” of Copyrighted Works, Associated Press (Jan. 10, 2024), https://apnews.com/article/openai-new-york-times-chatgpt-lawsuit-grisham-nyt-69f78c404ace42c0070fdfb9dd4caeb7 (observing trend of the courts in favoring technology companies regarding fair use of copyrighted material in AI). 

[21] Isaiah Poritz, OpenAI Faces Existential Threat in New York Times Copyright Suit, Bloomberg Law (Dec. 29, 2023) https://www.bloomberglaw.com/bloomberglawnews/ip-law/X1G6BBSO000000?bna_news_filter=ip-law#jcite (listing The Times’s requests for relief).