the adoption of ai in the legal advice may increase inequality instead of reducing it

Artificial intelligence (AI) can revolutionize multiple industries, work fields, and professions as it can automate almost any process and expedite time-consuming tasks. However, in the legal field, AI may only end up benefitting the rich and powerful, not the average citizen.

AI models need to be trained by using huge data sets. When it comes to generalist models like the popular ChatGPT and Claude, the developers behind them tend to rely on publicly available and free data sets to reduce the cost of training the software.

However, these models’ ability to provide accurate statements and analyses of legal information is limited due to the scarce data that they have been trained on. This prevents them from being useful, at least for now, to practitioners and customers who may rely on the advice of the AI model to build a case.

There is recent evidence that attorneys and customers have attempted to engage in a court proceeding and file paperwork with the assistance of ChatGPT and have presented erroneous materials.

Poor legal advice from an AI model is not necessarily an unexpected outcome for what should be considered a nascent technology. However, judges and court system administrators are worried that the widespread adoption of large language models (LLMs) could lead to more instances where practitioners and customers inadvertently rely on these solutions with unreal expectations and that could end up harming them from a legal standpoint.

To train field-specific AI models, specific data needs to be accessed, often at a not-so-negligible cost and that increases the cost and the ultimate price tag of such a service.

This is the case in the legal field, where case data including hearing materials, rulings, court procedures, evidence submissions, and other similar resources are often not available online.

Some businesses have created services that provide access to these materials but they are often quite expensive and that would propel the cost of training an AI model on these resources to the point that only wealthy customers may be able to pay for a subscription that gives them access to the technology.

A recent report from ARK Invest, a company that manages multiple tech and innovation-focused exchange-traded funds (ETFs), indicated that the cost of training AI models is declining rapidly at an annual rate of approximately 70%.

The study further asserts that “AI should increase the productivity of knowledge workers more than 4-fold by 2030. At 100% adoption, AI could increase global labor productivity ~$200 trillion, dwarfing the ~$32 trillion in total knowledge worker salaries”.

This may be true for fields that have huge publicly-available and free data sets on the worldwide web to train these models but may not apply to the legal profession where accessing court data is quite expensive and, in the case of some geographies, just an impossible endeavor.

Large companies have already identified the potential that AI models have to revolutionize the legal industry and are taking steps to acquire startups that have moved forward in that direction.

In February this year, the UK-based startup Robin.AI lured investors and raised $10.5 million in a funding round led by Plurar. The company developed a solution that relies on the LLMs created by Anthropic to help lawyers expedite the process of writing contracts and keep their library of legal resources properly sorted by tapping on the AI model as well.

Moreover, the data and media giant Thompson Reuters recently acquired Casetext, a company that creates AI-powered solutions for legal teams to write contracts, review legal documents, and analyze memos and court materials.

The startup was valued at $650 million and the deal was paid in cash. The size of the acquisition caught the attention of M&A, legal, and financial experts as it is a huge bet for Thompson Reuters and a statement of the potential that they see in the adoption of AI models in the legal field.

Whether those kinds of services translate into more affordable legal services is a hard estimation to make. The differences in how legal systems are administered worldwide and the level of access that software developers have to legal data – and its cost – will probably determine what the price tag will be and to which extent AI technology will benefit either the masses or the wealthiest when it comes to getting legal advice.