Note: This Essay was submitted to the TxP progress prize competition. You can find out more about TxP ('Technology and Policy') here.
I will make a modest proposal: with a few small reforms, we could make the English legal system significantly more competitive. I will use a few basic historical case-studies to show how Britain is unusually well-placed to take advantage of new technology in generative AI. I hope to make a novel case for both the importance of law to progress and Britain's comparative legal advantage.
Traditionally, progress studies have ignored law as a field of improvement. This isn't surprising -- economics treats law as 'transaction cost' which could ideally be removed altogether. It is treated essentially like friction in a rocket launch. Doing so, especially in the British context, is a mistake. Legal systems are essential components of all existing market economies, and in this regard English1 legal system is exceptionally successful. England has several of the world's largest law firms; a well-respected profession and judiciary; and receives a huge volume of business via choice-of-forum clauses in contracts. Legal services are one of the UK's largest service exports and constitute an important source of soft power. Nonetheless, competitive advantage is never assured and global law firms are in constant competition with each other and other dispute-resolution services.
Markets change. For the first time in 200 years a credible alternative to existing legal methods has been produced. I am referring to the varied suite of 'LawTech' products, which range from contract management software, document reviewers, general paperwork tools, to more ambitious proposals for low-cost advice generators. The market for these tools is already huge: legal tech companies are now valued in the billions and are used by the majority of law firms. Anyone with half a brain can see that new developments in large language models are relevant to this space. Law is a social technology which runs on natural language and is arguably the most textualized social practice on earth. In other words, legal systems contain a vast, highly complex textual corpus related to a single centralised discipline. When it comes to training legal models, these masses of text-as-data are basically natural resources each nation possesses exclusively for its own benefit:2 the crucial question now, as with all resources, is access.
My proposal is that all English case reports -- current and past -- should be made available to the public for free. Currently the majority of this immense corpus is controlled privately by two companies. First, the charity known as the 'Incorporated Council for Law Reporting' (ICLR); second, a company called the 'All England Law Reports.' Both charge high fees for access to these cases and restrict how they can be used. Neither are 'the' official reporter, but they nonetheless have an effective monopoly on the legal corpus. My proposal is to produce a public, machine-readable version of the law reporting corpus available to anyone who wants to use it. Why would this be beneficial? Simple: this corpus is one of the oldest continuous bodies of text dealing with legal cases and would be invaluable for creating LLM-powered legal tech.
Plurality is the name of the game: public access would allow researchers and innovators to train new (English) legal tools. For example, tools for giving legal advice and predicting court decisions, thereby vastly reducing the financial and time cost of legal services. This would both reduce transaction costs (which are considerable) and access to justice for all manner of clients. The success of these tools might also stimulate more extensive text-production, such as the publication of case reports at every level of adjudication and the immense mass of pre-litigation arguments. More ambitiously, if this technology reaches a sufficient level of reliability the government could offer it as an optional, alternative form of dispute resolution in civil cases. These developments would have the potential to make the English legal system into the most dynamic and advanced in the world.
England has many unique advantages when it comes to legal AI. First, its legal corpus. As a common law system, England has a rich and detailed record of previously decided cases. It is likely England has more total legal text than civil law countries which rely extensively on codes. England also compares favourably with other common legal systems. It has a culture of data storage (going back to the 12th century), pre-dates the others by hundreds of years, and is (legally) unbroken by revolution: 16th century cases are still cited in court. Consequently, England has a larger total corpus of useful, relevant material than virtually any other common law country.3 As with coal, England is richly endowed with the legal resources necessary for the next industrial revolution. Unlike coal, these resources are controlled by two actors which do virtually nothing4 with them.
Second, England has a tradition of market-oriented reform in the legal system. Historically, English judges and lawyers have regarded litigants as 'clients.' For example, in the 16th century there were three legal systems all operating side-by-side competing for client business. In the 19th century this service-provision mentality led to extensive court reforms (including the ICLR itself), such as the establishment of the county courts. This latter reform is notable for being choice-driven. It allowed litigants -- 'the customers' - to choose a single judge in the place of the jury. English lawyers still see themselves as service providers when it comes to private law: if you want AI-generated advice then you can have it.
Third, in England, much like the US, we accept data-mining of judgments. Civil law countries such as France have already banned the use of legal text-mining. One reason is because civil law countries typically pretend judges simply apply the law (found in codes) rather than making it with their own discretion. Text-mining threatens to undermine this fantasy by revealing judicial foibles. Conversely, English lawyers accept judges make law with their own personal political and moral beliefs. What does this all mean? In addition to possessing unparalleled raw textual resources, we also have right cultural framework to make use of them. Combine this with England's economic size, strong research base, and impressive legal market, and you get a clear national advantage.
It is rare for so much to be achievable with such a modest reform. Merely moving court reporting from a small private organisation to a national one would open up England's unique advantages in legal tech. So let's do it.
'English' because England and Scotland have separate legal systems. The former is substantially more economically successful than the latter.↩︎
Law is not like art. A model trained on French law is useless in England, and vice-versa. There is no plausible 'general model' with current legal texts.↩︎
The US is an exception, although the fragmentation into separate states means that there is no single corpus which can be used for a general tool.↩︎
The ICLR 'case genie', supposedly 'powered by AI', is remarkably crap.↩︎