Extraordinary headlines illuminating the latest and greatest application of the generative AI model ChatGPT come thick and fast. The newest version, GPT-4, can allegedly exhibit human-level performance and is the “most advanced system yet.”
Another provocative narrative of late is the multi-billion-dollar partnership Microsoft has fostered with OpenAI, the parent company of ChatGPT, with plans to integrate the tech into Microsoft's search engine, Bing.
Google also announced the upcoming release of their own generative AI equivalent Bard, signalling the start of what will inexorably become an overcrowded marketplace of large language models (LLMs).
Tech companies are sprinting at breakneck speed to cash in on the generative AI zeitgeist. Snapchat announced their My AI bot available to converse like a real friend. Amazon has been flooded with hundreds of ChatGPT authored books. WhatsApp is integrating ChatGPT to automatically craft responses. Spotify introduced their new robot DJ offering personalised commentary between songs.
But not every industry has been so quick to embrace the blockbuster bot with open arms.
Containing the Beast
Teaching is one obvious example of a profession that is seeing a seismic shift triggered by the emergence of ChatGPT, with opinion divided on how it should be used by students. The International Baccalaureate, an alternative qualification to A-Levels (or AP examinations in the US), recently permitted schoolchildren to quote ChatGPT in their content as a credible source.
Higher levels of education have not been as welcoming. Oxbridge are leading the charge in staunch opposition, gravely forewarning using ChatGPT constitutes academic misconduct. Others are more tolerant, with University College London celebrating the opportunity to teach students how to use emerging AI technologies ethically and transparently.
Waters remain murky when it comes to comprehensively explaining how ChatGPT will be properly policed in academia, with murmurings it may spell the end of traditional written coursework for good.
ChatGPT has ruffled feathers in all sectors and the legal sector is no exception.
Putting Theory into Practice
In the UK, Magic Circle firm Allen & Overy partnered with the mastermind behind ChatGPT, Open AI, to roll out their own chatbot (Harvey). The firm praised Harvey’s ability to work in multiple languages as a key deliverable in automating and enhancing a variety of legal tasks, ranging from contract analysis to regulatory compliance. In trials held last year, 3,500 of their lawyers asked the chatbot around 40,000 questions related to their day-to-day client work with “amazing results.”
Another British law firm, Mishcon de Reya LLP, is recruiting a GPT legal prompt engineer to help the firm understand how ChatGPT can be applied to their work. Daniel Hoadley, Mishcon’s head of data science, said:
Mishcon’s data science team has been working with LLMs in the practice of law space since 2019. We’ve seen our share of AI hype and underwhelming results over the past few years. ChatGPT and the recent release of the new GPT-4 model from OpenAI feels different. There’s a huge amount of work going on to understand how the firm can safely leverage these technologies to their fullest, and we rapidly implemented a policy banning the use of ChatGPT on client data and internal information while that work takes place. But I think it’s safe to say our people, including our lawyers, are now really paying attention to these technologies. We are particularly interested in using these models to give us an edge across a whole host of use cases, ranging from knowledge management to e-discovery, through to document drafting and legal research.
Across the pond in the United States, ChatGPT helped a design agency recover $109,000 from a client who ghosted them without spending a penny on legal fees. The bot also managed to pass prestigious law school exams at the University of Minnesota and even passed the multiple-choice segment of the US bar exam.
Generative AI isn’t just being experimented with in theory—it’s actually being deployed in legal practice. In Colombia, a judge openly admitted to having conversations with ChatGPT to inform his ruling when deciding whether an autistic child’s insurance should cover the cost of his medical treatment.
These developments have expanded the window of possibility for AI-powered judicial systems and sparked lively debates on what this means for the legal profession at large.
Friend or Foe?
ChatGPT burst onto the scene with a bang, dazzling legal professionals (along with the rest of the world) with its exceptionally futuristic capabilities. But after the initial fireworks, embers faded, excited chatter quietened, and with it came the slow, unsettling notion, that legal professions may have just met their successor.
Fortunately, the concept of ChatGPT completely usurping lawyers is an unlikely reality, but any role requiring written content will inevitably be affected by generative AI systems.
There are two perspective extremities when considering the application of ChatGPT. Legal teams need to land somewhere in the middle of viewing it as an existential threat and an indispensable necessity. Those who slam the door in the face of advancing technologies and are loathe to explore their potential, may find themselves eclipsed by competitors.
Like any lawyer worth their salt, a good starting ground is to evaluate the strengths and weaknesses in the arguments for and against allowing ChatGPT into their inner circle.
Weapon of Mass Construction
ChatGPT carries immense potential benefits for in-house counsels and law firms to tap into. Generative AI technologies akin to ChatGPT can streamline communications, lubricate protracted operations, and turbocharge productivity. Tech-savvy legal teams will have already noticed the potential cost gains and time savings up for grabs.
ChatGPT has the ability to accurately construct responses to legal questions in seconds. Drafting communications to stakeholders, clients, opposing counsel, and courts can cull time-intensive tasks and free up legal teams to focus on more strategically complex and valuable pursuits, such as prospecting new business or nourishing existing client relationships.
Rapidly generated responses help to facilitate fluid internal communications across departments and keep external third parties in the loop. Client queries aren’t left stagnant and instead, are promptly answered. This ensures cases can progress at a steady rate, pressures attached to meeting hard deadlines are diluted, and clients are continually reassured they remain unequivocally front of mind.
Automating administratively taxing and arduous tasks carry huge cost savings that empower law firms and in-house counsels to achieve more in less time. e-Discovery is one obvious example where generative AI models can drive cost and time savings by automating routine and tedious tasks. For example, during the privilege review process, Text IQ leverages its generative AI capabilities to suggest categories for privilege log creation, which can save 80 percent of the manual review time needed to create a privilege log.
Knowledge enrichment is another key benefit encapsulated in ChatGPT. Law firms and in-house counsel can ask complex questions that the AI will answer in natural language. This can be especially useful when composing summaries, verifying facts, or clarifying the intricacies of specific case law. The almost unfathomable corpus of data ChatGPT represents is an untapped treasure trove of information, instantaneously available at every lawyer’s fingertips.
ChatGPT can tangibly contribute to each stage of the e-discovery process, from reading, writing, researching, and planning. Providing additional insight into complicated subject matter and unshackling human resources chained to mundane document prep is a win for any legal department.
ChatGPT has proven its admirable level of competence and capacity to make life easier all round for humans—but those who embrace its use must proceed with caution.
Warning: Human Supervision Required
With great power comes great responsibility, and herein lies the catch with the all-singing-all-dancing wonderment of ChatGPT.
Far from a magic bullet, there are key limitations that must be considered carefully following its implementation—and that includes overestimating its capabilities. As far as its capacity goes in answering legal queries, ChatGPT so far has only been trained on data up to June 2021. This immediately paints an incomplete and dated legal picture.
Given the vast landscape of evolving regulatory and judicial parameters of each county, it is unlikely ChatGPT, in its current state, will have mastered the legal fluency required to keep pace with all legislative amendments internationally. This poses big problems for any lawyer looking to become solely reliant on ChatGPT as their singular source of truth.
To compensate for knowledge gaps, ChatGPT tends to “hallucinate” or fabricate answers. This presents a minefield of dangers for legal teams, as the AI does not clearly delineate where fact meets fiction. This is where human involvement becomes essential to perform thorough quality-control checks.
Another key area of concern is the scope for potential bias. ChatGPT only knows what it can glean from the data patterns it has been trained on. Therefore, any biases embedded in the knowledge base it was fed from, will permeate the information it churns out. Sexist, racist, homophobic, and xenophobic prejudices can unfairly skewer outputs and little clarity exists on how to effectively stamp out these inaccuracies.
Law firms and in-house counsel leveraging ChatGPT will also have to contend with issues of confidentiality—an area fraught with potential difficulties. Legal teams will have to carefully consider what sensitive data can be shared with AI to enhance operational efficiencies, whilst ensuring privileged information remains under lock and key.
Copyright infringement is another bone of contention, as lawyers will be unable to verify where ChatGPT gleaned its outputs from. If a lawyer uses ChatGPT to access information, and the AI generates a response that copies a significant portion of someone else’s authorship without permission, legal teams can find themselves in hot water.
Distributing third-party material without consent or failing to attribute the owner can mean those responsible for doing so are in the realm of serious copyright infringement. ChatGPT has even been likened in the industry to ‘’an essay with no footnotes’’—not an ideal scenario when meticulous accuracy is the fundamental underpinning of any competent legal department.
I’ll See You in Court!
The concept of AI-generated copyright infringement is already playing out for the world to see. In the UK, Getty Images have brought copyright infringement action against Stability AI, the developer of AI image generator Stable Diffusion, arguing millions of their photos were used to train AI without their permission.
In tandem, in the US, Microsoft, GitHub, and OpenAI have asked a judge to throw out a class-action lawsuit brought against them by a software developer who is claiming that the creation of the AI-powered coding assistant GitHub Copilot constitutes “software piracy on an unprecedented scale.”
The rulings in each case carry significant implications which will reverberate across the industry, as it will determine how AI is to be legally, morally, and ethically used going forward. The malleable concept of “fair use” is one judges will need to exercise caution in considering. Dependent on the direction the judiciary sways, it could open a Pandora’s box of never-ending AI copyright infringement claims.
The Final Verdict on Generative AI and the Law
The proactive use of ChatGPT in legal practice is a matter of when, not if. The fervour swirling around this revolutionary AI is laying the groundwork for meaningful discussion around how, why, and where advanced technologies should be used in the modern application of the law.
Although not a viable replacement for human lawyers, generative AI can be leveraged to great effect if the output is sufficiently scrutinised. AI should be harnessed to conduct the heavy lifting of research, eke out golden nuggets of evidence, and draft important initial documents. When deployed correctly, it’s a time-saving, efficiency-enhancing, invaluable game-changer for busy law firms and in-house counsel. As long as the material ChatGPT provides isn’t taken as gospel, lawyers can take advantage of their handy in-house ally to unlock new levels of productivity.
Legal teams can also protect themselves from copyright infringement accusations from third-party practitioners if they avoid copying material verbatim and instead re-write information to create originally authored content.
Legal teams should look to augment the skills of lawyers with the technical prowess of ChatGPT as a supplemental source of information. Clear ethical guidelines on the personal data it is trained on need to be established to avoid professional pitfalls and costly breaches of confidentiality.
In summary, ChatGPT has the potential to become a powerful legal tool that one should welcome into the fold. With careful supervision, reasonable expectations, and an open mind, lawyers with the willingness to embrace AI superpowers will be the ones to prosper in a world so increasingly saturated with utopian technologies.
Artwork for this post was created by DALL·E 2 based on a prompt from ChatGPT, under the direction of Kael Rose.