AI risks deepening unequal access to legal information

Receive free Artificial intelligence updates

We’ll send you a myFT Daily Digest email rounding up the latest Artificial intelligence news every morning.

The writer sits on the Civil Justice Council and the UK Ministry of Justice’s senior data governance panel

ChatGPT, the technology that launched a thousand quandaries, has now reached the attention of the most senior judges in the UK. This month, the House of Lords constitution committee heard evidence from the president of the Supreme Court. He described the experience of a litigant in Manchester who relied on ChatGPT for legal advice and inadvertently made an entirely fictitious submission to the court. Similar experiences have been reported in the US.

Concerns about generative AI models returning accurate legal information have led even ardent proponents of the technology to conclude that it should only be used to augment, rather than replace, the advice of lawyers. This is a pragmatic approach, albeit one that undermines its potential to expand access to legal advice to the growing numbers of people who are unable to afford it. However, in the rush to this solution, those in charge of the legal system are missing an opportunity to interrogate and address why the tool is returning inaccurate advice in the first place.

The root of the problem stems from the inability of the model to access sufficient relevant legal information. Experts in legal technology are clear that training generative AI on relevant documents, such as judgments and court decisions, can dramatically improve the accuracy of its responses. However, in England and Wales, access to these judgments and decisions is hard to come by. The most comprehensive collections belong to commercial publishers who charge large fees for access. The content held behind a paywall vastly outstrips that available to the public — a study in 2022 reported that only half of judicial review judgments are available on BAILII, a free website. 

Whilst the government has now invested in a service to make more information available, the decision to publish only those cases that judges deem to be significant and want online has led to issues. Research published in January 2023 found that only three-quarters of judgments given between May and July 2022 were available on the new, free to access website. Unless urgent action is taken, the asymmetry between the information available to those who can pay for it and that available to those who can not, is likely to persist indefinitely.

The lack of public access to a complete record of the judgments and decisions made by the courts has long been of concern to journalists and transparency advocates. The age of generative AI creates a new imperative to tackle the issue. If we do not, the best, most accurate and most powerful AI tools will be developed by and for those who already have access to legal information — private publishers, insurance companies and bulk litigants. This will exacerbate and entrench existing inequalities. 

Anyone doubting the advantage conferred by pre-existing access to legal data need only look to the recent acquisition of Casetext by Thomson Reuters. In June, Thomson Reuters purchased this relatively small company, which has only recently specialised in generative AI. On a call with shareholders, executives justified the size of the deal on the basis that, “they’re one of a very few sets of companies out there that also have access to data”. They clearly felt that the head-start afforded to Casetext by its ability to train ChatGPT4 on legal data warranted the $650mn pricetag.

Lawyers, judges and even the occasional justice minister are fond of citing clause 40 of the Magna Carta — “to no one will we sell, to no one will we deny or delay right or justice”. But while justice may not be for sale, access to the information generated by the courts most certainly is. In the age of AI, access to justice and access to legal information are becoming ever more closely intertwined. If we are serious about harnessing new technologies to address inequality in our legal system, addressing who gets to see and use legal data is where we need to start.