Artificial Intelligence in the Legal Field: Law Students’ Perspective

Daniela Andreeva¶,1 and Guergana Savova±, PhD

¶
London Metropolitan University, London, UK
±
Computational Health Informatics Program, Boston Children’s Hospital and Harvard Medical
School, Boston, MA

October 13, 2024

Abstract
The Artificial Intelligence (AI) field experienced a renaissance in the last few years across
various fields such as law, medicine, and finance. Firms have been devoting significant
resources to developing and evaluating AI applications including the creation of internal
AI Innovation Departments. While there are studies outlining the landscape of AI in the
legal field as well as surveys of the current AI efforts of law firms, to our knowledge there
has not been an investigation of the intersection of law students and AI. Such research is
critical to help ensure current law students are positioned to fully exploit this technology
as they embark on their legal careers but to also assist existing legal firms to better
leverage their AI skillset both operationally and in helping to formulate future legal
frameworks for regulating this technology across many industries. The study presented in
this paper addresses this gap – the intersection of law students and AI. Through a survey
of 27 questions conducted over the period of July 22 – Aug 19, 2024, the study covers
the law students’ background, AI usage, AI applications in the legal field, AI regulations
and open-ended comments to share opinions. The results from this study show the
uniqueness of law students as a distinct cohort. The results differ from the ones of
established law firms especially in AI engagement -- established legal professionals are
more engaged than law students. The law firm participants show much higher
enthusiasm about AI than this student cohort -- somewhat surprising as one would expect
the younger generation to be more open to new technologies. Collaborations with
Computer Science departments would further enhance the AI knowledge and experience
of law students in AI technologies such as prompt engineering, chain-of-thought
prompting, zero- and few-shot prompting, and language model hallucination
management. As future work, we would like to expand the study to include more
variables and a larger cohort more evenly distributed across locales. In addition, it would
be insightful to repeat the study with the current cohort in one year’s time to track how the
students’ viewpoints evolve.

Introduction
Since the late 2010’s the field of Artificial Intelligence (AI) has assertively re-emerged in the
public space due to the convergence of plentiful digital data, the advances in hardware designs
and capabilities, and scientific inventions in the representation of unstructured data including
text, images, video, speech, and music. The scientific inventions of representations such as the

1Corresponding author. LLB student at London Metropolitan University.

Email: [email protected]

1
transformer[1] and its flavors, operate on massive amounts of unstructured data to learn
language, facts and pixels enabled by unprecedented computational hardware that performs
mind-boggling matrix calculations. While the definition of AI and especially Artificial General
Intelligence (AGI) remains fluid in the current discourse of both academic and commercial
communities (for a review see Gebru and Torres[2]), from a practical point of view and for the
purposes of this paper, we adopt the operational definition of AI as a set of computational tools
for processing the unstructured data – mainly text and the rise of large language models
(generalist LLMs2, also referred to as foundation models) – to assist the legal professional in
their daily work-related activities. One might argue that these tools are predominantly from the
field of Natural Language Processing (NLP) – the field that endeavors to devise computational
methods for transforming, understanding and generating language -- because they deal with
classic NLP tasks such as information extraction, summarization, text generation, and question-
answering. Although the NLP field is related to AI, it is not absorbed by it; however, we leave
these nuances aside to focus on our investigation into the law students’ views, opinions and
concerns about AI as the students prepare for and grow into their careers.
A recent position paper by Kapoor et al[3] performs a careful analysis of the promises and
pitfalls of AI for legal applications and underscores its most important use case – information
processing, e.g. summarization, translation, e-discovery. Other applications discussed are
referred to as creativity, reasoning and judgment (e.g. drafting, automated mediation and
dispute resolution) and prediction (e.g. predicting court decision). The paper emphasizes the
importance of the involvement of legal professionals in the evaluation process of language
models to understand how the technology can best aid those in the field, a thesis also raised by
others (e.g. Cheong et al[4], Ma et al[5] Bhambhoria et al[6]). Most importantly, these studies
highlight cases where AI should not be used such as arguing court cases, decision making, and
judgments. These stances align with the AI TESCREAL philosophical framework developed by
Gebru and Torres where TESCREAL denotes transhumanism, Extropianism, singularitarianism,
(modern) cosmism, Rationalism, Effective Altruism, and longtermism3. Gebru and Torres urge
AI practitioners to work on tasks “for which we can develop safety protocols, rather than
attempting to build a presumably all-knowing system such as AGI”.
A 2024 white paper by Ma et al overviews the generative AI Legal landscape[5]. Generative AI
is a subclass of LLMs where the transformer method focuses on the preceding sequence to
generate the next target, e.g. the popular ChatGPT[7] family of models. As legal knowledge is
expressed in all types of unstructured form -- text, images, videos, audio -- the relevance of
these technologies is poignant. The authors review emerging technical solutions such as
retrieval-augmented generation (RAG)[8], prompt engineering[9], fine-tuning, alignment, and the
importance of legal datasets to train LLMs. They offer a legal technology market analysis across
five main segments of legal work: (1) research and analysis, (2) document
processing/contracting, (3) litigation, (4) time and billing, and (5) legal operations. The authors

2 LLMs refer to Large Language Models; not to be confused with LL.M. which stands for Master of Laws
degree. We use both LLM and LL.M. to refer to the relevant intended meaning. LLB or LL.B stands for
Legum Baccalaureus, Bachelor of Laws.
3 Following Gebru and Torres “we have capitalized some of the TESCREAL ideologies but not others. We

are following what members of these ideologies themselves capitalize: Extropians, Rationalists, and
Effective Altruists prefer to capitalize these terms, whereas transhumanists, singularitarians, cosmists,
and longtermists do not”.

2
estimate the legal technology market at ~13B USD in 2023 growing at 3.5% compound annual
growth rate. A separate study by Bloomberg lists the best current legal AI Tool options in the
legal tech market[10]. The Ma et al white paper raises the structural concern that “perhaps
underdiscussed, is the notion that the general workforce, particularly those who are young
lawyers entering the profession, may be ill prepared for the types of work that will result from the
automation of legal tasks by generative AI.”
A study by Mahari – a law student -- investigates the opinions and action items regarding AI
undertaken by law firms[11]. The study includes 59 senior and managing law firm partners
researching their views on and engagement with AI. It shows that legal practitioners have
embraced AI in the legal field and highlights the need for further practice-oriented research in
collaboration with NLP experts. Over 95% of the respondents state that their company is
engaged with AI in some form (discussions, conferences, training, etc.). Over 90% of the
participants express a positive attitude towards AI in the legal field (cautiously hopeful to
excited). Only a small percentage of the participants have little engagement with AI or a
concerned attitude towards the technology.
In parallel to these developments, the law student community has been increasingly active in
demonstrating their interest in AI applications of their future professional field. The UK-based
law firm Slaughter and May has been organizing Student Innovation competitions since
2020[12] drawing about 70 entries each year. The 2024 competition focused on generative AI
with the winning entry exploring the use of generative AI in Mergers and Acquisitions (M&A)
practice[13]. Other listed entries dealt with topics such as “What are the cultural challenges
with integrating generative AI into a law firm”, “Pick a practice area of your choice
(litigation was picked) and explain a potential use case for generative AI, setting out how
and why it will add value to the firm and its clients.” A blog post by a law student[14]
describes that “because of concerns about the accuracy and authenticity of generative AI only
28% of students would use tools like ChatGPT to help with their applications 4.”. Prof. David
Wilkins from Harvard Law School raises the question of how to best prepare students for
the new realities of legal careers[15]. There are an increasing number of blogs and
websites on the topics of law digital transformation geared to include early stage legal
professionals and academic viewpoints[16]. Topics cover technology developments as
well as policy implications[17] and regulatory frameworks, e.g. the EU AI Act[17].
Our study builds on the previous work and focuses on the intersection of AI and the future
generation of legal professionals, i.e. current law students. It contributes to the growing body of
investigations into AI in the legal field as it aims to take the pulse of the views, opinions,
awareness and uses of AI by law students in the legal field and potential areas of improvement
– complexity, accessibility, reliability, regulation and privacy being at the top. The paper is
organized as follows. We describe our methods, then present the results. The Discussion
section provides salient points for consideration. It is important to note that the attitude shared
by the authors of this paper is one of a cautious but non-limiting approach to the future and AI in
the legal field and far from the all-encompassing utopic views as defined – and critiqued -- in the
TESCREAL bundle[2].

4 Applications refer to the vacation scheme or training contract application forms and materials.

3
 Methods
We created a survey comprising of a list of 27 questions, split into 5 sections – Background (Q1-
3), AI Usage (Q4-12), AI Applications in the Legal Field (Q13-20), AI Regulations (Q21-25), and
Conclusion (Q26-27). The topics of the questions are listed in Table 1 in the Appendix. We used
Google Forms[18] to create the survey. We pilot tested the survey with three volunteers before
we distributed it broadly.
The survey aimed at a wide demographic of law students studying at a range of different
locations, universities and levels of studies. It was open for the period of July 22 – Aug 19,
2024.
We posted the survey on LinkedIn and in relevant LinkedIn groups. It reposted frequently to
ensure visibility. In addition, the survey was distributed via a student WhatsApp group as well as
direct outreach to law students. Personalized messages were sent to each participant,
explaining the aim of the study and providing details. Continued communication was established
with the participants to ensure clarity and engagement through which we ascertained their
student status. However, we did not formally vet the participants’ student status.
Participation in the study was voluntary and the data were securely stored within the internal
systems of our research team. The results were aggregated; therefore, confidentiality of the
participants was maintained.
The data were cleaned, structured and analyzed using Microsoft Excel[19]. Tables and figures
were created with the aggregated data to ensure the readability of the results.

Results
Table 1 in the Appendix lists the question topics from the survey and the breakdown per type of
answer. The survey attracted attention from undergraduate, graduate students and early stage
working professionals. There are 54 participants.
Survey Section -- Background (Q1-3) : The participants are predominantly undergraduate
law students (LLB students) (63%), 1 year (36%) followed by 3rd year (21%) and LL.M (26%)
st

and from the UK. The distribution shows participants at different levels, across the UK and some
outside of the UK (e.g. Nigeria, Pakistan, India).
Survey Section -- AI Usage (Q4-12) : Most of the participants state that they do not have
any AI training (91%), although they feel they have solid (54%) or good/user (43%)
understanding of AI applications but not necessarily the knowledge of the technical
underpinnings. This understanding is gained via the duration of their AI usage -- 75% indicate
usage of more than a month but less than two years, 10% had 2+ years of usage. Of the
participants who use AI applications, many of them use them multiple times per day (28%), with
18% of the participants indicating weekly usage of more than 5 times. The preferred tool is the
free ChatGPT 3.5 version[20] (65%) although some of them use both the free and the
subscription versions of ChatGPT[21] (20%). No participant states usage of open LLMs, e.g.
LLaMA 2/3[22] which is perhaps due to the lack of more user-friendly interfaces for these
models which in turn requires some level of computer science skills.
Furthermore, the participants actively use the AI tools to help with both or either school and
work (73%). They also use them for other tasks, e.g. planning and scheduling, financial

4
investments, cooking recipes, writing emails. Overall, they find AI helpful (98% spread across
“extremely helpful” 37%, “helpful” 27% and “somewhat helpful” 34%) and they do not indicate
any bias or prejudice in using it.
Survey Section -- AI Applications in the Legal Field (Q13-20) : Although the majority of the
participants find that AI applications are important in the Legal field (74%) and that they are
excited about them (50% spread across 20% excited and 30% mostly hopeful), only a small
portion of all participants (20%) indicated use of any of the more law-specific platforms. The
platforms the participants listed include Aria[23], Chatsonic[24], Harvey[25], ContractPodAi[26],
Luminance[27] and ChatGPT[7].
3% of the participants state that their university offers a formal training in AI in the Legal field as
part of the curriculum. The vast majority (86%) consider such training important. In addition, a
large portion (74%) are enthusiastic about collaborations with Computer Science departments
on applications of AI in the Legal field. The participants expressed the feeling that AI skills are
needed in the legal profession now (36%) and definitely in five years (51%) when most of the
survey participants will enter the workforce.
We left Q20 open-ended to allow recording of consequential applications of AI in the Legal field.
The participants were very active with their suggestions and enumerate 25 use cases. Among
them, the top ones are Drafting (especially contracts but also pleadings, judgments, general
statistics, and emails), summarization, legal research, information extraction and data science
(collection, extraction, tracking, data-driven decision making). Some participants felt that all
aspects of the Legal field could benefit from AI applications. Other use cases include case
findings, interpretation of judgments and precedents, judicial administration, marriage/deeds of
assignment, topic understanding, and proofreading.
Survey Section -- AI Regulations (21-25) : 38% of the participants feel that AI regulations do
not adequately keep up with the developments in the AI technology field, while 23% feel
otherwise. Most of the participants follow the EU AI Act[17] and the activities of the UK Office of
Artificial Intelligence[28]. However, only a small group (19%) follows current legal cases
involving AI, e.g. New York Times v OpenAI[29].
Survey Section -- Conclusion (Q26-27) : Q26 aimed to gauge participants’ interest in
participating in an AI and Commercial Awareness Law Student Club as a venue to keep up-to-
date of technology and legal developments in the area. The majority of the students are
interested in such a club. Q27 is open-ended asking to share any ideas or thoughts on the topic
of AI in the Legal Field. Twelve respondents shared opinions which we include in the Discussion
section.

Discussion
The results of this study demonstrate the uniqueness of law students as a distinct cohort when
researching AI in the legal field and the need for in-depth investigations to understand its
awareness and concerns. The Mahari study on generative AI in law firms included in the
Introduction section[11] show that 95% of the law firms are engaged with AI; compare that with
67% of our student participants indicating they feel excellently, well or somewhat prepared for
future usage of AI in the legal field. The law firm participants in the Mahari study[11] show much
higher level of enthusiasm about AI in the legal field compared to our student participants (90%
vs. 50%). This difference might be attributed to the level of experience in and exposure to

5
relevant AI tools between the participants in the two studies (senior/managing law firm partners
vs. law students). The two studies align in the expressed desire for interdisciplinary
collaborations between the legal and AI/computer science communities with Mahari listing
concrete initiatives that might be undertaken[11].
Our results and the differences with the Mahari study[11] point to the desire on the part of the
students for AI training. Although there are excellent LL.M.(see footnote 2 for abbreviation)
programs[30] focused on AI, they appear to be mainly at the Master’s level. Law student AI
training could be a collaborative effort with Computer Science departments which are well suited
to teach the technical underpinnings of the technologies including the intricacies of robust tool
evaluations and technology limitations (bottom-up perspective). Of course, such courses could
take many shapes[31], perhaps with minimal exposure to coding as it could be intimidating for
non-computational students and focus more on concepts and actual applications, e.g. prompt
engineering[9], chain-of-thought prompting[32], zero- and few-shot prompting[9], LLM
hallucination management[33] in the context of legal tools. Exposure to the AI philosophical,
regulatory and policy frameworks would enhance early-stage legal professionals’ thinking of
positioning AI in the broader context (top-down perspective). The convergence of the top-down
and bottom-up perspective of viewing AI would contribute to a cohesive vision. The learning
outcomes of such training would prepare the early-stage legal professionals not only for the use
of AI in the workspace and for leadership positions in the Innovation departments within their
firms, but also would prepare the next generation of jurists who will regulate AI as well as
represent and adjudicate AI-related cases.
Furthermore, none of the student cohort participants indicates experience with open models
such as Llama family of models[22] which have consistently shown performance results on
common evaluation benchmarks competitive with closed models such as ChatGPT. The use of
open models usually requires some level of programming skills. Therefore, as a call to open
model creators is to make available easy-to-use interfaces to stimulate their potential adoption
by early-stage legal professionals.
With a significant portion of the participants stating that there will be a need for AI skills in the
legal domain now (36%) and in 5 years (51%) the attention rightfully falls on what can be done
to expand the skillset of those already in the field or those who are law students (considering
that discussions mostly center around the future, but what about the now?). What individuals
can do is self-educate on understanding AI and machine learning basics, data science,
programming skills, prompt engineering and try to gain practical experience with AI systems and
legal AI systems. Online courses such as those offered on websites like Coursera[34] (on topics
like LLMs and prompt-engineering) could be beneficial. The overlap between the Bloomberg
best current legal AI Tool options referenced in the Introduction and the tools our survey
participants use consists of only HarveyAI and ChatGPT. This probably reflects the fact that the
majority of the participants are 1st year law students (36%), followed by 3rd year students (21%)
and LL.M students (26%) yet not fully exposed to these tools especially considering that most of
them require subscriptions by institutions. Another important aspect of self-education focuses on
ethics, security and privacy of AI in all aspects of the legal field to ensure the appropriate use of
the technology including for example scenarios such as the legal ramifications of a company
sharing employee data without consent with a company that builds LLMs. Familiarity with AI
interfaces and AI terminology would also prove beneficial as it would provide a practical and
easily applicable foundation for those seeking to expand their knowledge. For those aiming to

6
gain more in-depth knowledge, joint degrees between law and AI are available at some
universities.
Our study participants are very clear in their views on the applications of AI in the legal field –
73% of the respondents state using AI tools for help with school and work (together and
respectively) with the main areas of usage being text generation, summarization, legal research,
information extraction, data science and planning. Notably, very few utilize AI tools for nuance-
requiring tasks (e.g. arbitration) or accuracy-related tasks (e.g. referencing), and specific legal
domain tasks. This likely reflects the early stage of the participants’ career development but it
also highlights the limitations of the technology and provides valuable insights on the need for
further improvement and solutions for the full integration of AI in the legal field aligning with the
concerns expressed in Ma et al[5]. Accuracy and privacy appear as the primary challenges for
the broader adoption and integration of AI in the legal field especially for specialized legal
functions. Addressing the issue of accuracy, from a technology development point of view
combining LLMs with RAG systems could be the key solution to improving the accuracy and
relevance of LLM outputs by removing the need for additional training[5]. The RAG system will
provide real-time access to specific documents and data points which in turn will help with the
reduction of hallucinations and the improvement of the adaptability, diversity, explainability and
efficiency of the LLM. Open models (e.g. Llama) would provide a better alternative for those
looking for enhanced privacy by providing more transparency in data processing and handling
which in turn satisfies the growing demand for security within the legal field (and any high-
stakes field for that matter).
The technical complexities of the AI field are reflected in the current state of AI regulations and
the awareness of these regulations among law students. The data reveal that the majority of the
participants (63%) are not keeping up with AI regulations with the EU AI Act[17] being the most
followed. Additionally, 40% of the participants are unsure of whether regulations are keeping up
with innovation which shows the uncertainty among law students and the gaps to address.
There is a general uncertainty in how, when, where and how much AI usage is allowed and
acceptable within the legal field. Governments, employers and educational institutions need to
set clear guidelines and policies to make sure ambiguity is removed and keep the AI usage
process fair and ethical. A recent case[35] of the US-based law firm Levidow, Levidow &
Oberman exemplifies the dangers of naive AI usage[36]. Two of their lawyers used non-existent
citations in an aviation injury claim generated by ChatGPT believing the technology to be a
search engine and not aware of its tendency to hallucinate. A potential solution would be to
build an international united codex, applicable in every jurisdiction which clearly states the rules
around the usage of AI systems and ensure compliance.
There is a prevailing sense of concern and caution about the future of AI in the legal field. Our
study participants are concerned that AI might affect productivity and lead to intellectual laziness
and overreliance on the technology as well as a decline in job satisfaction. Moreover, the
concern reflects the rapid integration of AI and its potential to disrupt traditional legal practices.
However, the study participants remain hopeful about the future, if the technology gets mindfully
deployed without displacing human capital with the focus on streamlining mundane and time-
consuming tasks in the human-in-the-loop scenario. Furthermore, this cautious attitude offers
insights for educational institutions and companies to act in advance and prepare the workforce
for the evolving AI landscape through formal training (discussed above) and continued

7
education. The future of the legal field, to a degree, depends on fostering a balance between
innovation and preservation of human expertise.
Speaking of balance, it is important to mention that AI in the legal field follows the pattern of
disruptive technologies. The Innovator's Dilemma by Clayton M. Christensen[37] encapsulates
the essence of this ideology by explaining why most firms (or any institution for that matter) miss
out on new waves of innovation. No matter the industry, he says, a successful company with
established products will get pushed aside unless it knows how and when to abandon traditional
business practices. Christensen also leans into the pace of technological advancement and its
tendency to outstrip needs. Viewing AI in the legal field as a disruptive technology, we all falter
in how to approach it as it falls outside of our established patterns. This likely explains the
cautious attitude of our study participants – besides handling the load of their law studies, they
need to figure out what this new technology is all about and how that would affect their careers.
It should also be noted that Christensen’s model indicates that existing successful companies,
when faced with a disruptive technology, in general only survive by reinventing within; usually
through establishing a new internal group largely isolated from and unencumbered by current
business processes to grow and exploit the new technology. Overtime, this group typically
grows and ultimately consumes and replaces the old company. If this model holds for the legal
domain and AI, the opportunities for law students with training in AI and largely
“uncontaminated” by the existing legal models may be significant.
With an eye on the future, our plan is to initiate an AI and Commercial Awareness Law Student
Club. With 59% of the participants interested, the club is swiftly gaining popularity. The aim of
the club is to provide a venue to keep up with the developments of the AI and legal domain and
potentially initiate collaborations with Computer Science Departments on applications of AI in
the legal field. Another potential idea is to open a line of communication with university
administrations on creating AI courses for law students to prepare law students for the skills
needed for the world of the future. Additionally, the Club intends to regularly discuss the
evolving AI skillset with the goal of spreading awareness of the changing skillset needed for a
successful legal career in pace with AI developments.
Although this paper presents the first more in-depth study on the topic of law students’
perspective of AI in the legal field and includes more variables than previous work, it is not
without limitations. Our survey includes 27 questions answerable in about 20 minutes cognizant
of time limitations by the participants. Although the questions aim to cover the most important
variables, an expanded set of variables would be even more elucidating. In addition, the
majority of the survey participants ended up being from the UK. As future work, we would like to
expand the study to include a larger cohort more evenly distributed across locales. In addition, it
would be insightful to repeat the study with the current cohort in one year’s time to track how the
students’ viewpoints evolve.

Conclusion
While there are studies outlining the landscape of AI in the legal field as well as surveys of the
current AI efforts of law firms, to our knowledge there has not been an investigation of the
intersection of law students and AI. Such research is critical as the current law students are the
future legal professionals and a well-prepared workforce translates into not only operational
efficiencies for the firms but also into adequately prepared next generation of jurists who will
regulate AI as well as represent and adjudicate AI-related cases. This paper presented, to our

8
knowledge, the first study of this intersection of law students and AI. The results of the study
could serve as the basis to bridge the gap between AI knowledge and training on one hand, and
AI applications in the legal field on the other. Bridging the gap could take many forms:
educational institutions including AI training, governments working on the issue of
misinformation and building a universal AI codex, employers providing training and information
on AI to their employees from established experts, and individuals self-learning and acquiring
the AI skills needed for a modern legal career. Importantly, the results of this study demonstrate
the uniqueness of law students as a distinct cohort when researching AI in the legal field and
the need for in-depth investigations to understand law students’ AI awareness and concerns.

Acknowledgments
We are very grateful to the participants in the survey who enthusiastically volunteered their time.
We thank Prof. Sinha, London Metropolitan University.

References
[1]A. Vaswani et al., ‘Attention is All you Need’, in Advances in Neural Information Processing Systems,
Curran Associates, Inc., 2017. Accessed: Aug. 17, 2024. [Online]. Available:
https://proceedings.neurips.cc/paper_files/paper/2017/hash/3f5ee243547dee91fbd053c1c4a845aa-
Abstract.html
[2]T. Gebru and É. P. Torres, ‘The TESCREAL bundle: Eugenics and the promise of utopia through
artificial general intelligence’, First Monday, Apr. 2024, doi: 10.5210/fm.v29i4.13636.
[3]S. Kapoor, P. Henderson, and A. Narayanan, ‘Promises and pitfalls of artificial intelligence for legal
applications’, Jan. 10, 2024, arXiv: arXiv:2402.01656. Accessed: Aug. 22, 2024. [Online].
Available: http://arxiv.org/abs/2402.01656
[4]I. Cheong, K. Xia, K. J. K. Feng, Q. Z. Chen, and A. X. Zhang, ‘(A)I Am Not a Lawyer, But...:
Engaging Legal Experts towards Responsible LLM Policies for Legal Advice’, May 03, 2024,
arXiv: arXiv:2402.01864. Accessed: Aug. 22, 2024. [Online]. Available:
http://arxiv.org/abs/2402.01864
[5]Ma et al, ‘Generative AI Legal Landscape 2024’, Stanford Law School. Accessed: Aug. 27, 2024.
[Online]. Available: https://law.stanford.edu/publications/generative-ai-legal-landscape-2024-2/
[6]R. Bhambhoria, S. Dahan, J. Li, and X. Zhu, ‘Evaluating AI for Law: Bridging the Gap with Open-
Source Solutions’, Apr. 18, 2024, arXiv: arXiv:2404.12349. Accessed: Aug. 22, 2024. [Online].
Available: http://arxiv.org/abs/2404.12349
[7]‘ChatGPT | OpenAI’. Accessed: Aug. 27, 2024. [Online]. Available: https://openai.com/chatgpt/
[8]P. Lewis et al., ‘Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks’, Apr. 12,
2021, arXiv: arXiv:2005.11401. doi: 10.48550/arXiv.2005.11401.
[9]J. Wei et al., ‘Emergent Abilities of Large Language Models’, Oct. 26, 2022, arXiv:
arXiv:2206.07682. doi: 10.48550/arXiv.2206.07682.
[10] C. Dreyer, ‘Legal AI Tools: The 12 Best Options for Lawyers’, Rankings.io. Accessed: Aug. 27,
2024. [Online]. Available: https://rankings.io/blog/legal-ai-tools
[11] Mahari, Robert., ‘Bridging Disciplinary Disconnects: The Role of Legal Experts in Legal NLP’,
Network Law Review. Accessed: Aug. 19, 2024. [Online]. Available:
https://www.networklawreview.org/computational-four/
[12] ‘Innovation Competition’. Accessed: Aug. 27, 2024. [Online]. Available:
https://www.slaughterandmay.com/our-firm/innovation/innovation-competition/

9
[13] ‘Slaughter and May announces winner of 2024 student Innovation Competition’. Accessed: Aug.
27, 2024. [Online]. Available: https://www.slaughterandmay.com/news/slaughter-and-may-
announces-winner-of-2024-student-innovation-competition/
[14] ‘Levelling the Playing Field in Law 2024’. Accessed: Aug. 27, 2024. [Online]. Available:
https://www.cibyl.com//cibyl-insights/levelling-the-playing-field-in-law-2024
[15] J. Neal, ‘Harvard Law expert explains how AI may transform the legal profession in 2024’, Harvard
Law School. Accessed: Aug. 27, 2024. [Online]. Available: https://hls.harvard.edu/today/harvard-
law-expert-explains-how-ai-may-transform-the-legal-profession-in-2024/
[16] ‘Computational Legal Futures Archives’, Network Law Review. Accessed: Aug. 27, 2024. [Online].
Available: https://www.networklawreview.org/category/computational-legal-futures/
[17] ‘The AI Act Explorer | EU Artificial Intelligence Act’. Accessed: Aug. 27, 2024. [Online].
Available: https://artificialintelligenceact.eu/ai-act-explorer/
[18] Google. (2024). Google Forms (Version 1.0). Retrieved from https://www.google.com/forms/about/.
[19] Microsoft Excel 2016. Microsoft Corporation.
[20] ‘ChatGPT - Open A I Gpt 3.5’. Accessed: Aug. 19, 2024. [Online]. Available:
https://chatgpt.com/g/g-F00faAwkE-open-a-i-gpt-3-5?oai-dm=1
[21] ‘Introducing ChatGPT Plus’. Accessed: Aug. 19, 2024. [Online]. Available:
https://openai.com/index/chatgpt-plus/
[22] A. Dubey et al. 2024. ‘The Llama 3 Herd of Models’. Accessed: Aug. 27, 2024. [Online]. Available:
https://arxiv.org/abs/2407.21783
[23] ‘Aria browser AI | Free AI with real-time web access | Opera Browser’. Accessed: Aug. 19, 2024.
[Online]. Available: https://www.opera.com/features/aria
[24] ‘AI Chat Online for Free - AI Chatbot | Chatsonic: ChatGPT Alternative’. Accessed: Aug. 19, 2024.
[Online]. Available: https://writesonic.com/chat
[25] ‘Generative AI for Professional Services’, Harvey. Accessed: Aug. 19, 2024. [Online]. Available:
https://harvey.ai
[26] ‘Home’, ContractPodAi. Accessed: Aug. 19, 2024. [Online]. Available: https://contractpodai.com/
[27] L. T. Ltd, ‘Home Page’, Luminance. Accessed: Aug. 19, 2024. [Online]. Available:
https://www.luminance.com/https:/www.luminance.com/index.html
[28] ‘Office for Artificial Intelligence’, GOV.UK. Accessed: Aug. 19, 2024. [Online]. Available:
https://www.gov.uk/government/organisations/office-for-artificial-intelligence
[29] ‘NYT_Complaint_Dec2023.pdf’. Accessed: Aug. 27, 2024. [Online]. Available: https://nytco-
assets.nytimes.com/2023/12/NYT_Complaint_Dec2023.pdf
[30] ‘Top LL.M. Programs in Technology Law 2024’. Accessed: Aug. 27, 2024. [Online]. Available:
https://llm-guide.com/lists/top-llm-programs-in-technology-law
[31] R. Stross, ‘Computer Science for the Rest of Us’, The New York Times, Mar. 31, 2012. Accessed:
Aug. 27, 2024. [Online]. Available: https://www.nytimes.com/2012/04/01/business/computer-
science-for-non-majors-takes-many-forms.html
[32] J. Wei et al., ‘Chain-of-Thought Prompting Elicits Reasoning in Large Language Models’, Jan. 10,
2023, arXiv: arXiv:2201.11903. doi: 10.48550/arXiv.2201.11903.
[33] L. Huang et al., ‘A Survey on Hallucination in Large Language Models: Principles, Taxonomy,
Challenges, and Open Questions’, Nov. 09, 2023, arXiv: arXiv:2311.05232. doi:
10.48550/arXiv.2311.05232.
[34] ‘Coursera | Degrees, Certificates, & Free Online Courses’, Coursera. Accessed: Aug. 31, 2024.
[Online]. Available: https://www.coursera.org/
[35] F. K. Klein, T. Maulsby, R. C. Minkoff, and A. K. Alger, ‘MEMORANDUM OF LAW BY NON-
PARTIES STEPHEN A. SCHWARTZ AND LEVIDOW LEVIDOW & OBERMAN, P.C. IN
RESPONSE TO MAY 26, 2023 ORDER TO SHOW CAUSE’.
[36] D. Milmo, ‘Two US lawyers fined for submitting fake court citations from ChatGPT’, The
Guardian, Jun. 23, 2023. Accessed: Aug. 31, 2024. [Online]. Available:

10
 https://www.theguardian.com/technology/2023/jun/23/two-us-lawyers-fined-submitting-fake-court-
citations-chatgpt
[37] Clayton M. Christensen, The Innovator’s Dilemma: When New Technologies Cause Great Firms to
Fail. Boston, MA: Harvard Business School Press, 1997.

11
 Appendix
Q1: Type of student
LLB student 63%
MS 22%
Other (e.g. working at a law firm, lawyer) 9%
BA law student 6%
TOTAL 100%

Q2: Study year (for participants who indicated)

1st year 36%
Master's studies 26%
3rd year 21%
2nd year 15%
4th year 2%
TOTAL 100%

Q3: University name (for the participants who indicated)

London Metropolitan University 29%
ULaw 11%
BPP University 8%
University of London 5%
University of Warwick 5%
Àjàyí Crowther University 3%
Aston University 3%
BCU 3%
College of Legal Practice 3%
Hamdard School of Law, Pakistan 3%
National Law University, Delhi 3%
Stirling Univeristy 3%
UCFB 3%
University of Essex 3%
University of Exeter 3%
University of Linclon 3%
University of Manchester 3%
University of Plymouth 3%
University of Reading 3%
University of Southampton 3%
TOTAL 100%

Q4: Understanding of AI
Only as a user 43%
Solid but non-technical 41%
Solid with technical details 13%
In depth 2%
No understanding at all 2%
TOTAL 100%

Q5: Formal AI Training

No 91%
Yes 9%
TOTAL 100%

12
Q6: Usage of AI applications
Yes 70%
No 30%
TOTAL 100%

Q7: Length of usage of AI applications

More than a month but less than 6 months 28%
More than a year but less than 2 years 28%
More than 6 months but less than a year 20%
Less than a month 15%
2+ years 10%
TOTAL 100%

Q8: Frequency of usage of AI applications

Weekly, less than 5 times 38%
Daily multiple times 28%
Weekly, more than 5 times 18%
Other (ad-hoc basis, very rarely) 10%
Daily, less than 3 times 8%
TOTAL 100%

Q9: Types of AI applications used (for respondents with Yes for Q6)
ChatGPT 3.5 (free version) 65%
ChatGPT 3.5 (free version) and ChatGPT 4.0 (subscription) 20%
ChatGPT 4.0 (subscription) 15%
TOTAL 100%

Q10: Purpose of AI usage (for respondents with Yes to Q6: multiple answers)
To help with school and work 40%
To help with school 17%
To help with work 17%
Planning and scheduling 6%
Writing emails 4%
Cooking recipes 2%
Financial investment 2%
Job applications 2%
Other creative engagements 2%
Technical theory questions 2%
Thesaurus 2%
To get a summary 2%
To understand complex topics 2%
TOTAL 100%

Q11: Helpfulness of AI (for respondents with Yes to Q6)

Extremely helpful 37%
Somewhat helpful 34%
Helpful 27%
Not so helpful 2%
TOTAL 100%

13
Q12: Prejudice/bias against AI
No 80%
Yes 20%
TOTAL 100%

Q13: The importance of AI in the legal field

Extremely important 39%
Important 35%
Somewhat important 26%
TOTAL 100%

Q14: Attitude towards AI in the legal field

Cautiously hopeful 44%
Mostly hopeful 30%
Excited 20%
Extremely concerned/worried 6%
TOTAL 100%

Q15: Usage of AI applications in the legal field

No 80%
Yes (Aria, Chatsonic, Harvey, ContractPodAi, Luminance, ChatGPT) 20%
TOTAL 100%

Q16: Inclusion of AI in the legal ciriculum (multiple answers)

Very important 40%
Somewhat important 26%
Important 21%
Other 7%
My university already includes such training 3%
Not important 3%
TOTAL 100%

Q17: Collaboration with Computer Science department on applications in the legal field
Very enthusiastic 37%
Enthusiastic 37%
Somewhat enthusiastic 17%
Not interested 9%
TOTAL 100%

Q18: Level of preparation for the future usage of AI in the legal field
Somewhat prepared 46%
Not prepared 31%
Well prepared 11%
Excellently prepared 9%
N/A 2%
TOTAL 100%

14
Q19: Need for AI skills and background in the future of the legal domain
In 5 years 51%
Now 36%
In 10 years 8%
Never 6%
TOTAL 100%

Q20: Your opinion on impactful applications of AI in the legal field (multiple answers)
Drafting - contracts 20%
Summarization 17%
Legal research 10%
Information extraction 8%
Data science - collection, extraction, tracking, data-driven decision making 7%
Drafting - other, pleadings, judgements 7%
All aspects 6%
No idea/not sure 2%
Application forms 1%
Arbitration 1%
Automation - contracts 1%
Case finding 1%
Contract law 1%
Copyright 1%
Corporate emails 1%
Corporate law 1%
Criminal law 1%
Drafting - statistics 1%
Initial triage 1%
Interpretation of judgements, precedents, etc. 1%
Judicial administration 1%
Marriage/Deeds of assignment 1%
Mundane areas 1%
Topic understanding 1%
Visa generation 1%
TOTAL 100%

Q21: Keeping up with the landscape of AI regulations

Not really 63%
Somewhat 31%
Yes 6%
TOTAL 100%

Q22: Types of regulations followed (for respondents with Yes to Q21, multiple answers)
The EU AI Act 60%
The Office of Artificial Intelligence 40%
TOTAL 100%

Q23: Adequacy of the speed of innovation vs regulation catch up

Not sure 40%
No 38%
Yes 23%
TOTAL 100%

15
Q24: Following AI cases e.g. NewYork Times v Open AI
No 81%
Yes 19%
TOTAL 100%

Q25: AI's influence on opinions

Yes 65%
No 31%
Not sure 2%
Somewhat 2%
TOTAL 100%

Q26: Interest in AI and Commercial Awareness Law Student Club

Yes 59%
No 41%
TOTAL 100%

Table 1: Topics of survey questions and answer distribution, percentages are rounded. Note we list the
topic of the question, not the verbatim question. The sizes of the colored bars denote frequency, e.g. a
longer bar indicates bigger percentage. Background questions in blue. AI Usage questions in red. AI
Applications in the Legal Field in green. AI Regulations in yellow. Conclusion questions in purple.

16