By Dr Shahrul Mizan Ismail
There is something increasingly strange about legal education.
Law firms are integrating artificial intelligence into research, drafting, document review and knowledge management. Professional regulators are issuing guidance on how lawyers should use it responsibly. Courts are confronting AI-generated submissions and demanding greater human verification.
Yet some law schools still respond with a simple instruction: do not use it.
That position may feel safe. It is not. A law school cannot prohibit students from meaningfully engaging with AI while claiming to prepare them for contemporary legal practice.
This is not an argument for unrestricted AI use. It is an argument for something harder: teaching students how to exercise legal judgment when AI is available.
The professional direction is already becoming clear. The Malaysian Bar’s guidance on generative AI emphasises human oversight, verification and professional responsibility. The Solicitors Regulation Authority in England and Wales similarly warns that lawyers remain responsible for work produced with AI assistance. In the United States, courts are considering stronger requirements for human review of AI-assisted filings.
The message is not “let AI do the lawyer’s work”. It is: use the technology, but remain responsible for the law.
That is precisely why law schools must teach it.
The risks are not an argument against teaching AI. The risks are the curriculum.
Students need to learn why a convincing AI-generated case may not exist; when confidential information should never be entered into a system; how to verify authorities; how to distinguish organisation from legal reasoning; and how to recognise an answer that sounds sophisticated but is doctrinally wrong.
One does not acquire those abilities by being told never to touch the technology.
Law schools should instead create three learning environments.
First, an AI-free zone. Students must still demonstrate that they can read judgments, interpret statutes, identify legal principles and construct arguments unaided. Some foundational assessments should remain deliberately AI-free. Students cannot supervise reasoning they never learned to perform.
Second, an AI-assisted zone. Students may use AI but must disclose how. They might compare their interpretation of a case with an AI-generated interpretation, improve a draft argument and then explain which suggestions they rejected and why. The objective is no longer simply the polished final answer. It is to make judgment visible.
Third, an AI-required zone. Give students a difficult client problem and require them to use AI. Then assess the quality of their instructions, the mistakes they detect, the authorities they independently verify and the professional judgment they exercise before submitting the final work.
The sequence is simple: first, think without AI. Second, work with AI. Third, audit the AI.
This also changes assessment. For decades, legal education has largely assessed finished products: essays, memoranda, problem answers and examination scripts. Generative AI makes polished prose increasingly cheap. What becomes valuable instead is intellectual provenance.
How did the student reach the conclusion? Which authorities were checked? What did the AI suggest? What was wrong with it? What did the student change, and why?
Assessment must therefore move beyond “What is your answer?” and begin asking, “Show me your judgment.”
That is not lowering academic standards. It is raising them.
There will be discomfort. Some students will try to outsource their thinking. But prohibition can hide that problem rather than solve it. A student who secretly uses AI badly learns nothing about responsible use. A student required to expose, defend and critique that use can be taught something far more valuable.
The purpose of a law school has never been to protect students from difficult tools. It is to prepare them to exercise judgment in a difficult world.
Artificial intelligence has entered that world.
Law schools therefore face a choice: teach students to govern the technology, or send them into practice to discover its dangers by themselves.
They cannot indefinitely ban AI and still claim to produce AI-ready lawyers.
Associate Professor Dr Shahrul Mizan Ismail Faculty of Law Universiti Kebangsaan Malaysia (UKM)
