How ChatGPT and other generative AI may affect the legal industry


 

ChatGPT

 

A Brief Introduction to ChatGPT

 

“ChatGPT, or similar language models, has the potential to significantly impact the legal industry in various ways”, says ChatGPT when prompted, “how is ChatGPT going to affect the legal industry”. ChatGPT is a machine learning based technology created by OpenAI, an American artificial intelligence (“AI”) research lab aiming to make AI globally user-friendly and accessible.1 Since OpenAI’s launching of ChatGPT in November 2022, ChatGPT has become one of the most accessible and user-friendly AI models available to the public. Reaching 100 million users in just over two months, ChatGPT may be the fastest-growing consumer app in internet history, and there is no doubt about its influence on various fields such as finance, government, technology, etc.2

 

The legal industry has access to ChatGPT and other generative AIs at their disposal as well. Seeing a potential for success in the legal field, OpenAI released its latest version of ChatGPT called GPT-4 in March 2023. GPT-4 currently powers CoCounsel, the very first AI legal assistant program now widely used by the legal industry.3 GPT-4 can perform tasks typically handled by lawyers, such as reviewing documents, preparing for a deposition, conducting legal research, and summarizing documents.4 As AI advances at an unprecedented speed, its potential impact on the legal profession, especially for lawyers and paralegals, can be evaluated.

 

ChatGPT’s impact on the Legal Profession

 

a. What ChatGPT can do

 

Put simply, generative AI like ChatGPT collects a colossal amount of information and data, breaks it down into small units, analyzes these small units, and puts them together.5 Taking advantage of the generative AI’s ability to access and quickly access vast amounts of information, ChatGPT-4 can perform advanced tasks such as drafting documents, legal research, predictive analysis, contract review analysis, and due diligence.6 In fact, a LexisNexis survey from March 2022 revealed that over half of the New York lawyers who participated in the survey expressed that they had already used generative AI at work or were planning on doing so.7 Currently, the services in the legal profession most likely impacted by AI are those that involve routine and repetitive tasks which are mostly handled by junior associates, leading to cost savings for law firms as AIs can perform those tasks faster.8 A study done in March 2023, by researchers at Princeton University, University of Pennsylvania, and New York University, predicted that the industry most vulnerable to the new AI was legal services. 9 Another research by Goldman Sachs also predicted that 44 percent of legal work could be automated.10

 

The rise of generative AI could also affect the demand for legal services due to its affordability and accessibility. The LexisNexis survey revealed that of around 2,000 consumers in the legal market, around 15% of them stated that they had already tried generative AI for legal advice or assistance.11 As generative AIs become more and more advanced, the number of people turning to AIs for relatively simple legal services are likely to increase and attorneys will see a decrease in the number of potential clients for certain types of legal work.

 

b. What ChatGPT cannot do

 

Despite its enthusiastic reception and hype, ChatGPT as of now still remains more as a prospect, according to its inceptor; co-founder and CEO of OpenAI Sam Altman tweeted in December 2022 that ChatGPT is “incredibly limited but good enough at some things to create a misleading impression of greatness.”12

 

A perfect example of his comment comes from a lawyer who used ChatGPT to draft a brief. Most recently in May 2023, a New York lawyer used ChatGPT to draft a brief opposing a motion to dismiss filed against an in-flight personal injury case filed in the Southern District Court of New York.13 When the lawyer asked ChatGPT to search for relevant cases to include in his opposition, ChatGPT cited more than half a dozen cases involving flight injuries.14 The brief contained pinpoint citations from cases such as Martinez v. Delta Air Lines, Zicherman v. Korean Air Lines, and Varghese v. China Southern Airlines.15 However, it turned out that ChatGPT had invented those cases and the lawyer who did not think that ChatGPT could “make up” cases used them without confirming the validity of the cases. The lawyer was eventually penalized to pay $ 5,000 for sanctions under Federal Rule of Civil Procudure Rule 11 for submitting pleadings that contain arguments that have no evidentiary support.16

 

In addition to the aforementioned problem of ChatGPT “making up” facts, ChatGPT is also not up-to-date. ChatGPT does not update its database automatically and/or regularly. 17 The current version has only been tested with information up to 2021.18 Therefore, ChatGPT is not aware of and/or does not have any information after 2021, which means its collection of case law, statutes, local ordinances, legislature, news, etc. is outdated and may sometimes even be irrelevant. For example, ChatGPT does not have in its database most recent U.S. Supreme Court cases such as Dobbs v. Jackson Women’s Health Organization that overturned Roe v. Wade in June 2022 and Students for Fair Admissions, Inc. v. President and Fellows of Harvard College that struck down affirmative action in June 2023. Furthermore, because ChatGPT operates by processing potentially inaccurate information, it also produces inaccurate information, with OpenAI claiming a factual accuracy rate between 70-80%.19 In such a fast paced world where facts have become ever more important, outdatedness and inaccuracy can result in significant negative impact on users in various ways. This in turn means that it is still a little early for people to fully rely on the information and/or work product produced by generative AI assistance.

 

Legal Implications that Comes with the Increased Use of AI in the Legal Field

 

a. Privacy

 

Another concern with generative AI tools like ChatGPT is privacy. According to the OpenAI’s product and service privacy policy, “data ingested into the public ChatGPT model becomes part of the data repository and is not kept private,” and therefore the data could be breached and private information could be leaked.20 Hence there is a potential violation of a duty of confidentiality in using public generative AI, and many firms either ban or only allow limited use of generative AI in fear of privacy violation.21 Similarly, private users should keep in mind that the data they put in the public AI could be exposed.

 

AI

 

b. Copyright

 

As explained previously, generative AI such as ChatGPT creates text based on a vast amount of data that is already out, and a huge part of which is copyrighted. Courts’ ruling on whether ChatGPT’s use of such work constitutes an act of copying works that are protected by copyright is not clear because there has not been a case raising the issue. However, it is an implication that users, especially lawyers, need to keep in mind. On the other hand, whether work created by the generative AI can be copyrighted is another interesting question; there is not an abundance of cases regarding the issue. However, the U.S. Copyright Office in its formal guidance said that works created by AI may be copyrightable, provided that the work involves sufficient human authorship. 22

 

Verdict?

 

Artificial intelligence has more heavily been integrated in people’s everyday lives than they may have realized. From customer service, education, content creation, and businesses, more people are turning to tax-filing softwares such as Turbotax to file their tax returns and more and more websites are using AI chat-systems to direct and help customers. The advancement and development of AI are likely not ending anytime soon. In February 2023, Google launched Bard, its experimental artificial intelligence23 and Meta also introduced its own artificial intelligence, LLaMA (Large Language Model Meta AI).24 The tech industry’s seemingly infinite possibilities fuel competition and investments, and nobody can predict how fast and far it can take us. According to UBS, the AI services market is expected to grow to $90 billion by 2025.25 However, amidst the possibilities, the example of the New York lawyer who used ChatGPT to draft a brief with nonexistent cases reminds us that artificial intelligence is at its core “artificial.”

 

After a brief pause, ChatGPT added when asked “how is ChatGPT going to affect the legal industry”: “However, it’s important to note that while ChatGPT can be a useful tool, it should not be seen as a substitute for human legal expertise. Legal professionals will still play a crucial role in interpreting and applying the law, exercising judgment, and providing tailored advice to clients. Additionally, ethical considerations surrounding the use of AI in the legal industry, such as privacy, data security, and bias, need to be addressed to ensure responsible and fair deployment.”

 

1OpenAI,https://openai.com/ (last visited July 18, 2023).

 

2 Sawdah Bhaimiya, ChatGPT May Be the Fastest-Growing Consumer App in Internet History, Reaching 100 Million Users in Just Over 2 Months, UBS Report Says, Business Insider (Feb. 2, 2023), https://www.businessinsider.com/chatgpt-may-be-fastest-growing-app-in-history-ubs-study-2023-2.

 

3Casetext https://casetext.com/ (last visited July 18, 2023).

 

4ID.

 

5 Kevin Roose, How Does ChatGPT Really Work?, N.Y.Times (Mar. 28, 2023),https://www.nytimes.com/2023/03/28/technology/ai-chatbots-chatgpt-bing-bard-llm.html

 

6The Potential Impact of Generative AI on Law Firms, FairfaxAssociates.com (May 10, 2023), https://fairfaxassociates.com/insights/the-potential-impact-of-generative-ai-on-law-firms/

 

7Generative AI Captures Iagination of Lawyers, Law Students, Consumers Alike, LexisNexis.com (Mar. 20, 2023), https://www.lexisnexis.com/community/pressroom/b/news/posts/generative-ai-captures-imagination-of-lawyers-law-students-consumers-alike

 

8Id.

 

9Jan Hatzius, Joseph Briggs, Devesh Kodnani & Giovanni Pierdomenico, The Potentially Large Effects of Artificial Intelligence on Economic Growth, Goldman Sachs (Mar. 26, 2023), https://www.gspublishing.com/content/research/en/reports/2023/03/27/d64e052b-0f6e-45d7-967b-d7be35fabd16.html

 

10Ed Felton, Manav Raj & Robert Seamans, How will Language Modelers like ChatGPT Affect Occupations and Industries?, Social Science Research Network (Mar. 6, 2023), https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4375268.

 

11Id.

 

12 Sam Altman (@sama), Twitter (Dec. 10, 2022 7:11 PM), https://twitter.com/sama/status/1601731295792414720?lang=en.

 

13Benjamin Weiser & Nate Schweber, The ChatGPT Lawyer Explains Himself, N.Y.Times (June 8. 2023), https://www.nytimes.com/2023/06/08/nyregion/lawyer-chatgpt-sanctions.html

 

14Id.

 

15 Benjamin Weiser, Here’s What Happens When Your Lawyer Uses ChatGPT, N.Y.Times (May 27, 2023),https://www.nytimes.com/2023/05/27/nyregion/avianca-airline-lawsuit-chatgpt.html.

 

16 Mata v. Avianca, Inc., No. 54 Civ. 1461 (S.D.N.Y. June 22, 2023), available at https://storage.courtlistener.com/recap/gov.uscourts.nysd.575368/gov.uscourts.nysd.575368.54.0_3.pdf; see also Debra Cassens Weiss, Lawyers Who ‘Doubled Down’ and Defended ChatGPT’s Fake Cases Must Pay $5K, Judge Says, ABA Journal (June 26, 2023), https://www.abajournal.com/web/article/lawyers-who-doubled-down-and-defended-chatgpts-fake-cases-must-pay-5k-judge-says.

 

17 OpenAI, What is ChatGPT?, https://help.openai.com/en/articles/6783457-what-is-chatgpt (last visited July 18, 2023).

 

18Id.

 

19OpenAI, GPT-4, https://openai.com/research/gpt-4 (last visited July 18, 2023).

 

20Skye Witley, ChatGPT Tempts Big Law Despite AI Accuracy, Privacy Worries (2), Bloomberg Law (June 1, 2023), https://news.bloomberglaw.com/privacy-and-data-security/chatgpt-tempts-big-law-despite-ai-accuracy-data-privacy-worries.

 

21 Stephanie Pacheco, ANALYSIS: AI Has Entered the Chat- Is the Legal Industry Ready?, Bloomberg Law (May 10, 2023),https://news.bloomberglaw.com/bloomberg-law-analysis/analysis-ai-has-entered-the-chat-is-the-legal-industry-ready.

 

22Copyright Registration Guidance: Works Containing Material Generated by Artificial Intelligence, 88 Fed. Reg. 16, 190 (Mar. 16, 2023) (to be codified at 37 C.F.R. § 202); see also Blake Brittain, AI-created Images Lose U.S. Copyrights in Test for New Technology, Reuters (Feb. 22, 2023), https://www.reuters.com/legal/ai-created-images-lose-us-copyrights-test-new-technology-2023-02-22/.

 

23 Sundar Pichai, An Important Next Step on Our AI Journey, Google (Feb. 6, 2023), https://blog.google/technology/ai/bard-google-ai-search-updates/.

 

24 Introducing LLaMA: A foundational, 65-billion-parameter large language model, MetaAI (Feb. 24, 2023),https://ai.facebook.com/blog/large-language-model-llama-meta-ai/.

 

25UBS Editorial Team, Let’s Chat about ChatGPT, UBS (Feb. 23, 2023),https://www.ubs.com/global/en/wealth-management/our-approach/marketnews/article.1585717.html.

 



The Copyright Implications of AI-Generated Art


Today we are witnessing the rapid advancement of artificial intelligence (AI) technologies which are capable of generating human-like images, audio, and text.1 While many people are excited by these AI technologies, these advancements have also rung the warning bell for many individuals in creative industries who believe that AI technologies are exploiting their work and harming their profession as a whole. This is especially a concern for the art industry

 

Recently, AI technologies which generate images have become increasingly sophisticated. These AI technologies are programmed to learn how to generate images from scraping publicly accessible data – in this case images – from the internet.2 Consequently, artists are becoming concerned about how their artwork online is being exploited by large companies to train their AI to create new works, often for a profit.

 

Last year, artists around the world took to the internet and reposted images of a red prohibition circle over the letters “AI” in protest of image-generating AI. The first person to post this image appears to be an artist named Alexander Nanichtkov who stated in a tweet that “AI creates the ‘art’ you see on the backs of artists being exploited . . . AI ‘art’ is currently scraping the web for art and uses it in datasets. No artist gave consent to have their art used. We were not compensated.” 3 In the face of these concerns, many legal professionals are currently considering the extent to which intellectual property rights – particularly copyrights – serve to protect the work of artists.

 

AI Generated

 

A variation of the symbol posted in protest of AI generated art.4

 

Under the Copyright Act, copyright protection is only granted to “original works of authorship.” 5While there is no law or provision of the United States Constitution which defines who may be considered an “author,” the UnitedStates Copyright Office (USCO) tends to only recognize copyrights for works “created by a human being.” 6 Courts actually have a history of denying copyright protection to non-human authors. For instance, in Naruto v. Slater the Ninth Circuit Court of Appeals held that Naruto, a crested macaque, did not have legal standing to claim copyright infringement under the Copyright Act for pictures that Naruto took 7 himself. And for those wondering, Naruto’s claims were filed by the People for the Ethical Treatment of Animals (PETA).8

 

Beyond the context of copyright protection for animals, courts have also considered cases regarding copyright protection for AI systems.This past April, the Supreme Court declined to hear an appeal from a computer scientist named Stephen Thaler, who filed a suit to claim a copyright on behalf of hisAI system, known as Device for Autonomous Bootstrapping of Unified Sentience (DABUS).9 Although Thaler argued that DABUS created the works autonomously, the Court upheld the decision of the lower courts and the USCO, and declined his appeal “on the grounds that the AI couldn’t be considered the legal creator of those works.” While these decisions may seem promising to human artists since they deny copyright protection to non-human authors, the issue of AI generating images is far from settled.

 

The U.S. Patent and Trademark Office acknowledges when these AI technologies scrape data from the internet to generate images, this process “will almost by definition involve the reproduction of entire works or substantial portions thereof,”10 and as a result, copyright infringement seems quite plausible. In fact, a recent decision by the USCO has ruled that AI generated works – whether that is art, writing, or music – will not be granted copyright protection. 11 This decision arose from a copyright application for a comic called Zarya of the Dawn, which used original writing alongside art generated by Midjourney, an AI program.12 In their letter to Zarya creator Kristina Kashtanova, the USCO explained how “the office will not register works produced by a machine or mere mechanical intervention from a human author” and how the crucial question is “whether the ‘work’ is basically one of human authorship, with the computer . . .merely being an assisting instrument, or whether the traditional elements of authorship in the work . . . were actually conceived and executed not by man but by a machine.” 13

 

book

 

The cover page and second page of Zarya of the Dawn14

 

On the other hand, OpenAI, a company that uses AI generating tools, has argued that the works created by AI tools should be protected since they qualify as fair use since the process of scraping data from the internet is done to “create a useful generative AI system and the copies aren’t made available to the public.” 15 To provide another recent example, this past February Getty Images sued Stability AI alleging that it “copied at least 12 copyrighted images from Getty Images’websites” to train their Stable Diffusion AI program.16 While Stability AI responded with a fair use defense, Getty argues that this defense is inapplicable since Stability AI’s AIprogram undermines the market for Getty’s copyrighted material. At the same time, Stability AI is also facing a class action lawsuit from several artists alleging copyright infringement due to the use of their images to train their AI programs.

 

AI image-generating tools have also caused controversies since they can scrape data to mimic a particular artist’s style.This past January, DeviantArt – an online art website – was sued for copyright infringement after they began offering a new service in which users could pay a monthly subscription to access an AI art generator.17 This AI art generator was trained on the artwork of artists who uploaded their art to DeviantArt for free,and users could enter a text prompt to generate images.18 Furthermore, users of this AI tool could even input a specific artist’s name in order to create an image which replicated that exact artist’s style.19 The claim argues that this constitutes copying and consequently breaches copyright. 20 While this may seem like a clearcase of copyright infringement, not everyone in the legal community agrees. For example, Andres Guadamuz, a legal school at the University of Sussex, believes that these AI tools are simply learning patterns from the original works, brushstrokes, and styles which are not covered by copyright law.21

 

To add another layer of complexity to this matter, the USCO has stated that an AI-generated work could be copyrightable if an individual can prove that “they themselves put a meaningful amount of creative effort into the final content.” 22 In other words, while a company that uses an AI system, which scrapes images from the internet to produce new content, may be found to have committed copyright infringement, if that company can prove they added ameaningful amount of creative effort to the final product, then the company could actually argue that they have a valid copyright in the new work.23 This would overcome the previous problem that individuals have faced when they tried to make an AI system a copyright holder.

 

This does not mean however, that minimal contributions made to an AI-generated work would automatically qualify for copyright protection. The Director of the USCO Shira Perlmutter stated “If a work’s traditional elements of authorship were produced by a machine, the work lacks human authorship and theOffice will not register it.”24 For example, if an AI system produced a work solely based on a human prompt then the “’traditional elements of authorship’ are determined and executed by the technology – not the human user.”25 On the other hand, if an AI-system produced a work of art based on a human prompt,and then that finished work was then edited further using Photoshop, the USCO has stated that a copyright is more likely to be granted.26

 

This new stance on the copyrightability of AI-generated works brings with it a wide array of new problems. For example,there is no definitive standard or test which could be used to determine whether a company or individual has made a meaningful enough contribution to an AI-generated work which would qualify it for copyright protection. As previously mentioned, the Director of the USCO Shira Perlmutter stated that if a meaningful human contribution is made to an AI-generated image, through Photoshop for example, then the AI-generated work could qualify for copyright protection.However, there would need to be further clarification or case law on how much Photoshop editing would be required to qualify as a meaningful contribution to the work. Companies could quickly make contributions to an AI-generated work on Photoshop and as a result they could defeat copyright infringement claims.This would be especially harmful to artists whose original works are used without their consent to train these AI systems which then create new works fora profit. Additionally, these new works created by AI systems could be sold at a far cheaper price since producing them is drastically quicker than the cost of human artists creating original works. This could lead to severe economic harms to human artists and could have even greater implications for theart industry as a whole.

 

Despite the bleakness of this situation, there may be a light at the end of the tunnel which could prevent generative-AI from gettingout of hand and inflicting severe harm on human creators. One popular method which has been discussed to control the rapid growth of generative-AI is theimplementation of some sort of licensing system.27 This type of licensing system would require companies using generative-AI to pay copyright holders a fee forusing their works in training their AI systems. This licensing system could offset some of the harm caused by generative-AI since human artists could refuse to license their original works, or they could at least be compensated for allowing their work to be used to train an AI system. Unfortunately, this licensing systemmay not address the greater problem of unfair competition between generative-AI and human artists since companies could produce new works at a far quicker and cheaper rate than human artists creating original works.

 

Due to the recent advancement of AI technologies and their benefits, it is likely that generative-AI systems will continue to be used for the foreseeable future. Furthermore, since these AI technologies are very new, our laws and courts are not fully prepared to handle the quickevolution of these new technologies. As a result, the consequences and effects of generative-AI systems must be continually considered since they can pose substantial harm to human artists.

 

1 Eric Revell, AI complicates copyright law, YAHOO! FINANCE (May 19, 2023), https://finance.yahoo.com/news/ai-complicates-copyright-law-131932670.html.

 

2Id.

 

3 Butlerian Jihad, ARTISTS MASS PROTEST AGAINST AI STEALING THEIR WORK BY SHARING ANTI-AI LOGO, THE_BYTE (Dec. 16, 2022),. https://futurism.com/the-byte/artists-protest-ai

 

4Verity Babbs, Digital Artists Are Pushing Back Against AI, HYPERALLERGIC (Mar. 6, 2023), https://hyperallergic.com/806026/digital-artists-are-pushing-back-against-ai/.

 

5 17 U.S.C. § 102

 

6Revell, supra note 1.

 

7Naruto v. Slater, 888 F.3d 418, 426 (9th Cir. 2018).

 

8Revell, supra note 1.

 

9Id.

 

10Id.

 

11 Sam Sachs, US Copyright Office Rules AI-generated artwork, content not legally protected, WFLA (Feb. 23, 2023, 2:30 PM),https://www.wfla.com/news/national/us-copyright-office-rules-ai-generated-artwork-content-not-legally-protected/.

 

12Id.

 

13Id.

 

14Richard Lawler, The US Copyright Office says you can’t copyright Midjourney AI-generated images, THE VERGE (Feb. 22, 2023, 9:06 PM),https://www.theverge.com/2023/2/22/23611278/midjourney-ai-copyright-office-kristina-kashtanova.

 

15Id.

 

16Id.

 

17 Darian Woods & Adrian Ma, AI-generated images breach copyright law, artists say, NPR (Feb. 7, 2023), https://www.npr.org/2023/02/07/1155185861/ai-generated-images-breach-copyright-law-artists-say.

 

18Id.

 

19Id.

 

20Id.

 

21Id.

 

22Katyanna Quach, AI-generated art can be copyrighted, say US officials – with a catch, THE REGISTER (Mar. 16, 2023), https://www.theregister.com/2023/03/16/ai_art_copyright_usco/.

 

23Id.

 

24Id.

 

25Id.

 

26Id.

 

27Kai Nicol-Schwarz & Tim Smith, Why Harry Potter is the copyright timebomb under generative AI models, SIFTED (May 18, 2023), https://sifted.eu/articles/generative-ai-copyright.