A quiet revolution is underway in courts and law firms from New York to Nairobi. Generated artificial intelligence (AI) with large-scale language models like GPT-4 is no longer a distant concept or novelty in Silicon Valley. A few years ago, they restructured research automation, research automation, drafting streamlines, and legal decision-making support.
On platforms such as elite global law firms such as Allen & Overy, and Cocounsel from Case Text (which was recently acquired by Thomson Reuters), AI is already acting as a digital co-pilot. A few minutes to review contracts, prepare depositions, and generate notes. In response, regulatory frameworks such as the EU AI Act have introduced strict guidelines, particularly for high-risk use cases such as judicial decisions. Courts around the world have adopted careful optimism. Accepts AI efficiency while protecting against bias, opaqueness and dehumanization. Pakistan is currently in this arena.
In its April-April 2025 decision, SC Justice Syed Mansoor Ali Shah has announced an 18-page ruling that could define the trajectory of legal AI in the country. While arguing on seemingly routine rent disputes, the Supreme Court addressed a deeper issue. How can AI help fulfill Article 10 (the right to fair trial) and Article 37(d) (cheap and quick justice) of the Constitution? The ruling called AI “a viable tool for access to timely justice” and urged the National Judiciary (Policy Commission) to establish clear guidelines. These must ensure that AI remains a “promoting tool” and do not replace judicial discretion and do not compromise the conscience of human judgment.
Judge Shah's vision is coordinated. He highlighted non-negotiable boundaries, recognizing AI's promises in clearing the judicial backlog and democratizing access to legal knowledge. “Justice is human effort,” he pointed out. Algorithms can support, but only humans can weigh morality, emotion, and fairness. The ruling warns against biases built into historical legal data and AI's “hastique” tendencies. The court must employ AI, but it must be ethical and intentional.
The ground practitioners took action while the apex court deliberated on the principles. In May 2025, Pakistan saw its first generative AI masterclass for lawyers, a historic, practice-oriented training jointly organized by Enablyifyi and the Legal Aid Association. This two-night masterclass was not another theoretical webinar. It was a practical leap into the future.
Participants, including supporters, internal counsels and young legal minds learned how to use AI for rapid engineering, contract deficits, drafting case briefs, and summarizing decisions. In real time, they helped AI monitor draft legal provisions, analyze case law, and build customized legal tools.
One such production volume was the “Danish legal assistant” and the codeless AI paralegal capable of drafting citations and notifications to conduct both English and Urdu citations, notifications of Draft and notifications, structural discussions and legal research.
As Muhammad Shahzar Ilahi, a master class instructor and co-founder of Enablyai, stated, “Attentions who use AI will not take over, but perhaps those who do not use AI.” This statement reflects the broader truth. AI is not a replacement for experts. They amplify them.
Pakistan's legal system is overloaded, with over 2 million cases pending across the courts. The courts face the heaviest backlog. AI can assist in triage cases, automate repeat orders, recommend precedents, and even dramatically reduce turnaround times. The Federal Academy of Justice has already piloted the “Judge GPT,” which was developed along with ETH Zurich and deployed to more than 1,500 judges. Early reviews suggest faster legal research and improved quality of decision-making.
Legal services are very expensive for most citizens. AI can reduce the time to draft and review lawyers, allowing for more affordable legal services, especially for underserved populations.
But the risk is real. If AI systems are trained in biased judgments, they can perpetuate historical discrimination. If not checked, hallucinations can mislead both judges and lawyers. There is also the “black box” dilemma. How can the party challenge the decision if the AI logic remains opaque?
The answer lies in ethical infrastructure. The court must be liable. Human surveillance must be absolute. Transparency in AI usage is essential.
Pakistan now has to turn this initial momentum into structured reforms. Based on international best practices and national insights, several policy procedures are important.
Judicial AI Guidelines: As sought by the Supreme Court, state agencies must urgently define where AI can support (such as research, drafting) and where it should not intervene (judgment or verdict).
AI literacy in legal education: Bar associations and universities need to incorporate AI literacy into ongoing legal education and the LLB curriculum. The 2025 Masterclass is just the beginning. It needs to be expanded nationwide.
Localized Legal AI Model: Pakistan needs to build its own legal model trained in local laws, languages and judgments that are not dependent on foreign AI trained in American law. This ensures cultural and jurisdictional relevance.
Essential AI Disclosure Code: Judges and lawyers must disclose where AI contributed to research or drafting. This maintains professional accountability.
National AI & Justice Surveillance Forum: A multi-stakeholder watchdog made up of judges, engineers and ethicists, must oversee the legal integration of AI and address misuse.
The challenge is not whether AI will enter Pakistan's legal system. I already have it. The challenge is to shape that entry with wisdom, attention and constitutional clarity. If we succeed, Pakistan can become a global model of Ai-Aigmented justice where tradition and technology coexist in the public interest.
Justice must always evolve, but as we stand in this cusp of transformation, legitimate fraternity must guide us.
The writer is a defender of the High Court.
