In 2019, a metal cart rode in the aisle of an Avianca flight from El Salvador in New York and hit Roberto Mata on the knee. Mata has kept two lawyers for injury to continue the airline, which led to a case that made a legal precedent, but not for a reason for a reason, Mata or his lawyers would have hoped. Instead, the two lawyers have reached fame In 2023, when the judge of the case sentenced a fine to the pair $ 5,000 for invoking six fictitious cases, invented by artificial intelligence software, in short to the court.
Punishment was the first of its kind and was largely reported in legal circles as a prudence tale for lawyers on the risk of counting on AI. This edifying tale was not taken into account. During the two years that followed the incident, there is no less than 139 instances Legal memories containing quotes invented by AI, including more than 30 in May. Although many of them come from self-represented litigants, they also include memories deposited by eminent national law firms. And, in a case, a judge struck Expert testimony submitted by a scientific data of the giant anthropic AI – on the reasons where it contained a “hallucination” created by the own anthropic software.
These stories of AI were wrong in the courtroom are a lively source of gossip around the watercolors of the law firm. However, these cases are also aberrant values. In reality, the vast majority of lawyers using AI do so in a responsible manner, including verifying that the quotes produced by AI are really verified before putting them before a judge.
Many are also delighted with its usefulness. Lawyers say that AI excels in the research and revision of documents and that, more recently, they have found it capable of more difficult tasks such as the writing of memories. And all this progress has great implications for both the future trajectory of legal careers and for the law company.
A growing suite of AI tools
Most lawyers appreciated the transformative power of the AI at the same time as everyone: November 2022. It was at this moment that Openai published the initial version of Chatgpt to the general public, and, without surprise, it was about six months later that the first wave of false quotes began to appear in the audience rooms.
In fact, the idea of using AI to enlighten legal research has existed for more than a decade. Pioneers in the field include Ironclad and Disco, which respectively offer automated means of creating and managing contracts and accelerating the discovery process.
The difference today is that, as in other industries, the last three years have seen the legal profession acquire and deploy AI technology at a larger and faster pace than ever before, in particular because the generative AI has become more easily available and more powerful.
In February, the Federal Bar Association published an survey of more than 2,800 law professionals: 21% said they used AI in their practice in 2024, and the number was much higher – 39% – for lawyers for large law firms. A separate lawyer’s survey published in April by the Transcription Service revise found a much higher overall adoption, 48% of respondents claiming to have used AI for research.
Unsurprisingly, the number of AI offers aimed at the legal profession increases. There are services like Harvey, which is built on AI models with companies such as Openai, Anthropic and Google and specializes in writing and research; And Clio based in Vancouver, British Columbia, which offers practice management tools. Clio raised $ 900 million in 2024, evaluating the company at $ 3 billion; At the end of June, Harvey raised $ 300 million in a financing round that values the company at $ 5 billion.
For the moment, there are no clear data on the way lawyers deploy an AI in their work use these specialized tools, or if they are rather on popular LLMs created by Openai and Google, which all have ingested trains of statutes and common law decisions.
According to Mark Lemley, an intellectual property teacher from Stanford well known in legal technology, most law students and junior lawyers probably use general services like Chatgpt, because they are the most accessible.
In large companies, a popular AI tool is Coconsel, which was developed by the legal startup Casetext. The Thomson Reuters data giant bought Casetext in 2023 and now offers Coconsel as part of Westlaw, a popular but expensive service widely used by so-called lawyers.
Daryl Landy, litigant of the national company Morgan Lewis, says that he likes CoSONel in part because he is designed to take into account the special obligations of the legal profession concerning customer confidentiality and data protection. In practice, this means that AI invites created by Landy and its colleagues are not part of a training set which can enlighten the search for people outside of its business – an important question in a profession in which practitioners often have access to very sensitive information on their customers.
Modification of professional standards
For decades, most law firms are perfectly written in the “Finders, Minders and Grinders” model. The sentence describes a pyramidal structure where a handful of lies at the top bring customers and put most of the work to a second level of higher lawyers, which in turn supervise a larger pool of associates.
Where is an AI agent in the pyramid? It has been obvious for some time that AI has been a good choice to move large stretches of work of banal “crusher” – and indeed this occurs with tasks such as the examination of documents, the drafting of contracts and basic research. It is a good thing in many ways. The AI already eliminates part of the chore of legal work, a large part of which is traditionally in the search for billable hours – a notorious aspect of the profession which often lets customers wonder how many hours on an invoice represent occupied work rather than essential tasks.
With this in mind, it is difficult to challenge a technology that considerably reduces the time devoted to tedious tasks, reduces costs and promises to release lawyers for deeper and more strategic questions. However, there are drawbacks, especially with regard to the training of the next generation of lawyers.
“A way you get good legal instincts for logical arguments is to obtain representatives – by treating things and doing them again and again,” explains Lemley. He fears that the wider adoption of AI means that young lawyers could not acquire some of the analytical skills that have traditionally been acquired thanks to an in -depth reading of case law and statutes.
Lemley also observes that AI is increasingly assuming tasks such as writing memories and craft arguments, which many lawyers consider the funny part of work, and which generally fall to more senior lawyers.
All this raises the question not only of how lawyers will spend their time in the future, but also how they will charge their services. For years, customers have fought – generally with limited success – to introduce alternative arrangements in order to escape to pay on time for Junior Associates to lower legal rabbit holes.
In the future, says Lemley, the situation can be reversed as customers – realizing that AI can perform many legal tasks in a few hours and not hours – to be invoiced for fixed time blocks, while law firms offer flat costs and other arrangements.
The legal profession has long been able to keep its business model, so that such changes are probably not on the immediate horizon. But for the moment, it becomes clear that AI already provides lawyers something precious: additional time.
According to the Thomson Reuters survey, the widespread adoption of the AI will probably release four hours a week for the typical lawyer, or 200 hours per year – hours that could be used to extend their customers, develop additional skills or, of course, stack billable hours.
A primacy of law for AI lawyers
Pablo Arredondo, who has created the popular Coconsel AI tool and is now vice-president of Thomson Reuters, frequently organizes seminars for judges on the intersection of law and artificial intelligence.
During his work, he often reflects on an almost centenary sentence which appears in the very first section of the federal rules of the civil procedure – namely that the judges should encourage the resolution “fair, fast and inexpensive” of each action.
“These three words are not how someone would describe disputes to us,” said Arredondo.
This could change if the courts adopt AI tools in a general and effective way. And, in fact, such adoption already begins, especially in Massachusetts where the judicial system adopts automated tools to help tenants confronted with the expulsion to navigate in complex legal forms.
“As this technology can be used for better results, there is a moral imperative to use it,” explains Arredondo. However, he warns that any deployment of AI by courts and lawyers must occur with the strictest surveillance.
Arredondo is not alone in this feeling, because the legal profession, like so many other industries, tries to balance the efficiency gains offered by the AI with security. In 2023, reflecting on how technology is ready to disrupt the work of judges, none other than chief judge John Roberts wrote: “Any use of AI requires prudence and humility.”
