AI contract review is no longer a test case. It is real. Legal teams across India now use it daily. The global AI contract review market crossed $465 million in 2024. It grows at 26.5% per year. India alone has 662 active legal tech firms. They have raised $793 million in total. And 74% of in-house counsel say they want AI tools to flag risks and draft contracts.

The speed gains are hard to ignore. In one study, an AI tool reviewed five NDAs in 26 seconds. Lawyers took 92 minutes for the same task. The AI scored 94% on issue spotting. The lawyers scored 85%.

But fast is not the same as safe. And in India, that gap can cost you.

Why Indian Law Demands More from AI

India's contract law is not a single system. It is a layered one. Central laws, state rules, and sector norms all apply at once. The same contract may need to clear FEMA, DPDP, a state stamp act, and a sector body each with its own terms.

This is why getting a contract right at the drafting stage matters more in India than in most markets. Once a dispute arises, it costs time, money, and senior leadership focus. Catching risks at the review stage is the smartest way to manage legal risk. And that means the AI reading the contract must know Indian law and not just guess at it.

Not all AI is built the same way, and the output reflects that. Stanford research found that off-the-shelf LLMs get legal queries wrong 58% to 82% of the time. Legal AI trained on primary sources and paired with human review performs at a very different level. The point is not whether AI works for legal review. It does. The point is that design and training data set apart AI that is ready for Indian law from AI that is not.

Five Things Any AI Must Get Right for Indian Contracts

Stamp duty varies by state. Each Indian state sets its own rates. Under Section 35 of the Indian Stamp Act, 1899, a contract that is not stamped or not stamped enough cannot be used as proof in court. The fine for short stamping can go up to ten times the gap. In 2023, the Hon'ble Supreme Court ruled in N.N. Global v. Indo Unique Flame that even a dispute clause in an unstamped contract cannot be enforced. Any AI that does not check for state-wise stamp duty misses this risk at the door.

FEMA applies to cross-border deals. Any contract tied to foreign funds or cross-border payments must follow the Foreign Exchange Management Act, 1999 and RBI rules. A missing RBI approval clause or a pricing term that breaks capital account norms can draw fines of up to three times the sum at stake. If an AI review skips FEMA checks, it is not complete for Indian use.

The DPDP Act, 2023 has changed data contracts. India's Digital Personal Data Protection Act, 2023 in partial effect since November 2025, fully live by May 2027 holds data owners liable for what their vendors do. Fines go up to Rs 250 crore. Data Processing Agreements are now among the most high-stakes parts of any service contract. The DPDP Act has its own distinct needs: a consent manager setup, a Data Protection Officer based in India, and a framework that does not split personal data into "sensitive" and "not sensitive" the way other global laws do.

Indian laws on lawyer duty apply to AI use too. Under Section 35 of the Advocates Act, 1961, lawyers owe a duty of strict secrecy. The Kerala High Court said in 2025 that judges must check all AI results with care. The Bombay Bar Association put out its own rules the same year. No Indian law yet speaks to AI use in legal work. But the duties are clear: if an AI tool leaks client data or gives a wrong answer that a lawyer signs off on, the fault lies with the lawyer. Not the tool.

Language matters by law. The DPDP Act, 2023 says terms and notices must be offered in all 22 languages listed in the Eighth Schedule of the Constitution. Contracts with workers, gig staff, or buyers in local languages carry legal force. Most AI tools only read English.

What Makes a Review Defensible

The word "defensible" has a clear meaning here. It means the review process can hold up if someone tests it in court, in an audit, or before a board.

Three things make a review defensible in India.

First, playbooks built for Indian law. Not off-the-shelf clause banks. Playbooks that code in stamp duty rules by state. FEMA triggers. DPDP duties. These must be part of the core review logic not add-ons.

Second, AI that shows its work. Each flag, each edit, each output must be logged. Under Section 65B of the Indian Evidence Act, 1872, digital records such as AI logs need proper sign-off to count as proof in court. If a contract dispute goes to trial, the firm must show what the AI flagged, what the lawyer checked, and what was cleared. The audit trail is not a nice feature. It is a legal need.

Third, a human must always sign off. Cyril Amarchand Mangaldas, India's largest law firm, says it sees a 20–30% speed gain using AI for due diligence. A global firm pilot with C3.ai was 95% correct and cut review time by over 80%. But in each known case, the final call stayed with a human lawyer. The ABA's advice is simple: treat AI like a first-year lawyer. Check the facts, the sources, the logic, and the law before you approve.

The Real Stakes

India's legal landscape is among the most complex in the world. Many rule-makers. State-level rules. Evolving laws like the DPDP Act, 2023. The margin for error in contracts is narrow. And the cost of getting it wrong is high.

Getting a contract right the first time is not about saving hours. It is about sound risk management. A contract that fails on stamp duty, misses a FEMA rule, or lacks DPDP-ready data terms does not just create a legal issue. It creates one that costs far more to fix than it would have cost to prevent.

AI can help. But only if it is built and used with the care that Indian law demands. The true test of a good AI review is not how fast it runs. It is whether the output holds up when it matters most.