Tuesday, June 18, 2024

AI and the law

While artificial intelligence modules are slowly taking over many positions within the Indian administrative system, the Indian justice system is so complex that AI may only have peripheral applications in this field

By Sujit Bhar

When Justice SA Bobde was the chief justice of India, he had strongly pushed for the adoption of Artificial Intelligence (AI)-based IT models that could ease and support the administration of justice in India. An app was released that could lead to the translation of Supreme Court judgments in more than nine regional languages. Today, the AI-based software being used is called SUVAS (Supreme Court Vidhik Anuvaad Software), which translates legal papers from English into vernacular languages and vice versa.

So far, no major complaint has come to the fore about this software’s functioning, but one has to remember that the scope of this software is limited. What could happen if and when different AI-based IT models are broadly pressed into service within the Indian justice system?

The Indian judicial system is immensely complex, cumbersome and inertia-driven. The question of widely introducing AI in the Indian system might end up like any attempt to introduce driverless vehicles on Indian roads. The number of variables in India is so large that it will be difficult for even an advanced, self-learning software to fathom its depth.

Even in the US, where the application of laws is more the norm than the exception, strange incidents have been reported involving AI. Recently, in that country, a fiasco was just about averted. And it wasn’t an isolated incident. There has been a series of pretty high profile AI-generated legal briefs citing fake cases and misstating facts. Michael Cohen, Donald Trump’s former lawyer, asked this model—Generative AI (this is supposed to create texts and images based on prompts)—for case references. The response was supposed to be humanlike. Instead, the software spewed out completely fake cases, created by the software. The lawyer unintentionally included these phony cases generated by AI in a brief last month, and this was only detected when the opposing attorney cross-checked the case references and found nothing.

When you realise that one day humans—which would include lawyers, court officers and even judges—might become so dependent on AI-generated documents that even this simple cross-checking will not be done, you understand the grave danger inherent in this.

US Supreme Court Chief Justice John G Roberts Jr, however, sounds optimistic. In an article, he recently accepted that legal research may soon be “unimaginable” without AI, though he also said: “I predict that human judges will be around for a while.”

He added in the note: “AI obviously has great potential to dramatically increase access to key information for lawyers and non-lawyers alike, but just as obviously it risks invading privacy interests and dehumanizing the law.”

If those statements sound paradoxical, they probably should be, simply because there is yet to be a fixed point analysis of AI-driven systems. Such systems learn and grow, which means that the opinion or analysis it will present tomorrow will probably be different from what it gave today. Would that be a good point of reference for any judge or lawyer?

The pressing concern

At the other end of the spectrum, there is a massive problem waiting to be tackled within the Indian judicial system. It’s the mountain of pendency which extends to all levels of courts in the country, from taluka courts to the Supreme Court. It is not just about the common man being denied his/her rights of being able to access justice in India, but such incredible levels of pendency ultimately reflects very poorly on the ability of the republic to protect its subjects.

With the advent and fast spread of social media, an age of instant justice has developed, slowly moving the faith of the people away from the courts. The common man expects the courts to be able to mete out justice within a reasonable time frame. This feeling of mistrust in Indian courts also rubs off on potential international parties, wanting to invest in India or even just wanting to pay the country a visit. When the rule of law is compromised, the country itself withers.

Sacrificed within this disastrous system are high calibre judges who could have made a real difference in a disturbed world. Even if it is possible to increase the number of judges and the courts exponentially, the already existing pendency may never be fully addressed, given the rise in new cases from the population explosion in the country.

Within this atmosphere lies the possible scope of AI. The present modules may come up short at every turn. However, if such advanced models are given the opportunity to learn the nuances and intricacies of India’s judicial system, would it be possible for such AI modules to look into the pendency in a new light and offer ways to solve the problem?

Hurdles galore

What are the areas in which artificial intelligence can have an immediate effect? The first would be the back reference section, where it can scan mountains of case files and suggest the best possible matches. Secondly, the modules could also be used for drafting, such as business contracts, lease agreements, adoption papers etc.

Trouble may start even when they are used for standard court procedures, such as granting of default bail and of parole (other than in UAPA or PMLA, etc cases). Even the granting of police custody or judicial custody could have come into the ambit of AI, but the “established” Indian “system” might defeat any such attempt.

Consider the case of a bail petition made by a repeat offender. Under normal circumstances the investigating officer (his lawyer) would resist bail. However, if the AI module wants to strictly abide by the 90-day automatic bail principle (and this has been supported by statements made by a number of Sup­reme Court judges), then there is a problem.

Consider the case of parole. Two issues are possible. First, if the criminal, say, wants to go and attend the cremation of a dead parent, a judge could grant this on humanitarian grounds. Secondly, if it is found that this criminal has tried a jailbreak earlier, then the judge, on the advice of the IO and/or the warden, could disallow bail.

How would any AI-driven module ascertain the difference?

Lack of online documentation

Despite the continual exhortations of Supreme Court judges over the last few years, barely a sentence of orders and documents in tehsil courts are available online. If the AI-driven module is to learn the ways of the Indian judicial system, where would it get the input? Inputs only from High Courts and the Supreme Court would mean data from mostly appeal cases that have seen their germination in lower courts. Much of the original records could be lost in transcript or may remain in parts. This is very possible in cases that remained stagnant for years.

Hence pendency issues could pose bigger problems than new cases, primarily because of the lack of proper documentation in those instances, and also because laws themselves have gone through major changes over the years.

Added efficiency

AI modules, though, can definitely be a tool to help lawyers and judges become more effective in their approach, reducing time. AI modules are designed to rationalise the decision-making process by summarising all relevant information in a more efficient way than the human brain is able to do.

This happens through cross referencing with earlier judgments and legal options. The paralegal would probably one day become redundant.

Finally, like in every field, AI would be able to take over a lot of the work with the justice system, though the famous Indian “system” would make it nearly impossible for such modules to ever replace the existing set of professionals within the system.


News Update