AI and Law: Trends and Tendencies

Dr Vinod Surana
Managing Partner & CEO, Surana & Surana International Attorneys

I am delighted and honoured to speak before this august gathering on AI and Law which is a subject of great interest to me. I will place some random thoughts and observations before you with the hope of instigating serious contemplation.

Today’s context is set in the backdrop of the work of eminent scholar Mr. Richard Susskind – famous for over three decades of work and publishing on the impact of technology on the professions. Prof. Susskind in his latest book predicts the decline in today’s professions and describes the people and systems that will replace them. The basis of his claim is that current professions are antiquated and no longer affordable. He believes that most legal functions and jobs will be automated. He lays great stress on technology taking over.

On the other hand, Prof. David Maisser of Harvard Law School makes a point based on empirical evidence, that the keys to success for individuals and corporations are passion, people and principles. The only competitive advantage is from the ability of skilled managers/professionals to create and transfer enthusiasm, excitement and determination and this will be directly proportional to customer satisfaction and profits. Herbert Marcus makes a point that technology is more than a technical apparatus. It is a social process where humans are inseparably involved.

Drawing on these streams of thought, we can identify some general observations and challenges: technology has always (i) magnified human nature, desires ambitions etc. (ii) reduced cost
(iii) reduced effort (iv) simplified tasks (v) improved efficiency (vi)

reduced waste (vii) led to concentration of wealth and power. In the light of this, how might AI, the latest permutation of technological change, affect the practice of law?

Early last year when ChatGPT was introduced my firm did a month-long experiment on the use of ChatGPT in the practice of law and we came up with some very interesting findings:

Negatives:

1. Privacy issues – these need to be addressed by developers of the model. There is uncertainty on the privacy safeguards adopted by the large language model (LLM), hence all sensitive/confidential data input during the research was redacted and/or deleted.
2. Authenticity – with respect to information such as case law; since the database is general in nature and non-specific, the relevant case law information extracted is often not reliable.
3. Bias – due to the inbuilt system of relying on the most popular views and the majority opinion, the AI tool is likely to show its bias echoing the above views. So, user caution is always advised.
4. Echo chamber – Another drawback of LLM AI is that they work in an environment or ecosystem in which their beliefs are amplified or reinforced by communication and repetition inside a closed system and insulated from rebuttal.

Positives:

1. Legal notices – Drafting legal notices based on raw instructions received from the client over call/email or text messages. The results were 90% accurate.
2. Reply to legal notices – Drafting reply notices to legal notices received. It is able to adopt the style of writing/drafting when prompted but requires supervision in relation to proof reading and cross-verification with the instructions provided to us. The results were 75% accurate.
3. Identifying relevant legislation – AI is able to do a very good job in extracting the various provisions/ rules/ regulations/circulars/etc., discussed in any given judgement. It is also able to identify the various laws that pertain to a proposition and to prepare a concise note on the same when prompted. The results are 95% accurate.
4. Explanation of legal concepts – The system is clear on the fundamentals of a concept. There is in fact a greater degree of clarity and simplicity in the explanations offered by the system than there is in most legal literature available for reference.
5. Framing of arguments – ChatGPT for the most part can understand the content in a prompt and is able to frame arguments based on the concepts therein.
6. Questions for cross examination – ChatGPT is able to draft decent possible cross-examination questions. In some areas it has drafted questions to trap witnesses as well. For those with little knowledge or experience with cross-examination, it is a very useful tool.
7. Input for targeted arguments – in relation to drafting detailed written arguments and submissions based on the pleadings, including the Plaint, Written Statement, Evidence etc, ChatGPT can be used to frame arguments where specific, detailed and pointed commands / prompts are given.

Artificial Intelligence represents a great opportunity for the profession to leapfrog. Entry barriers such as high-quality support, infrastructure etc have fallen dramatically. Young lawyers with the ability to think calmly and work systematically will be able to match most seniors in spite of their decades of experience.