AI-Generated Contracts and Their Legal Risks

Under the Contracts Act 1950, the validity of a contract rests on four familiar pillars: offer, acceptance, consideration, and intention to create legal relations.

The law is indifferent to authorship. Whether the words come from a lawyer or an artificial intelligence (“AI”) tool, the legal question remains the same: is there a binding agreement? 

But that neutrality should not be read as reassurance. A contract does not become more reliable simply because AI produced it. If anything, the growing use of AI tools introduces risks that are easy to miss, precisely because the output so often looks like it has everything in order.

Fluency can disguise fragility 

One of the more deceptive qualities of AI-generated contracts is their surface credibility. The output tends to be well structured, formally worded, and persuasive in tone. The kind of document that gives you confidence before you have even read it properly. 

That is part of the problem. The language may read well, but the substance underneath could still be off. Clauses might pull in different directions, important details may have been left out, or the contract might borrow legal concepts from another country that simply do not apply under Malaysian law. 

This matters because Malaysian contract law operates within a specific statutory framework. The Contracts Act 1950 governs the formation and enforceability of contracts but it does not stand alone. Depending on the nature of the agreement, other legislation such as the Sale of Goods Act 1957, the Consumer Protection Act 1999, the Employment Act 1955, or sector-specific regulations may also apply. An AI tool trained predominantly on foreign legal material may produce clauses that are inconsistent with these local statutes, or that inadvertently exclude protections the law implies by default.

This happens because AI does not actually understand what it is writing. It generates text by recognising patterns from large amounts of existing material. It has no knowledge of the specific transaction, no awareness of the commercial context, and no ability to anticipate how a clause might be interpreted if a dispute arises.

Unlike a lawyer, AI carries no duty of care for the advice embedded in its output. If the clause fails, there is no professional accountability to trace back to the tool that produced it. The risk lands squarely with the party who chose to rely on it. 

Uncertainty is enough to undo the deal

AI generates text by recognising patterns, not by understanding the law. It does not know whether a clause will hold up in court, whether it conflicts with local legislation, or whether a judge would even be able to make sense of it.

Section 30 of the Contracts Act 1950 is straightforward: if terms are not certain or are not capable of being made certain, the agreement is void. It does not matter how neatly it was formatted.

This provision has real teeth. Malaysian courts have consistently held that vague or incomplete terms cannot be saved by good intentions or subsequent conduct alone. If the contract does not clearly record what the parties agreed, the court may decline to enforce it at all.

AI tools are particularly prone to producing clauses that look complete but are legally thin. They may use standard boilerplate that does not fit the specific deal, leave gaps where the parties’ actual intentions should have been recorded, or create ambiguity by combining phrases in ways that make grammatical sense but no legal sense.

A contract built on that kind of foundation may never be tested, but if it is, the consequences can be significant. Voided agreements, unenforceable obligations, and disputed terms are not outcomes that get resolved cheaply or quickly.

Confidentiality risk is happening in the background

There is a second category of risk that tends to receive less attention, one that does not arise from what the contract says, but from how it was created.

When parties use AI tools to draft contracts, they routinely input sensitive information without pausing to consider where it goes. That information may include the identity of the parties, commercially sensitive terms, or personal data. Once submitted to an AI tool, that information may be processed, stored, or used in ways that fall outside the control of either party.

From a legal standpoint, this raises several concerns. Inputting a counterparty’s information into an AI tool without proper consent from the parties involved could constitute a breach of the Personal Data Protection Act (“PDPA”), even if the intent was simply to get a contract drafted quickly.

Beyond the PDPA, employees who feed commercially sensitive information into AI tools may be breaching implied duties of confidentiality owed to their employer, or express obligations under their employment contracts. Where a non-disclosure agreement is already in place between the contracting parties, sharing covered information with an AI tool could trigger a breach of that non-disclosure agreement and expose the disclosing party to a claim.

Responsibility Still Has a Human Face

AI is changing how contracts are written. But it is not changing what the law expects. 

In the eyes of Malaysian law, responsibility still has a human face. The party who signs the contract, the lawyer who advises on it, the business that relies on it, these are the actors the law holds to account. 

No court will apportion blame to a chatbot, and no statute recognises an AI tool as a party to an agreement. When something goes wrong, the law looks for a person. 

Technology moves fast. Legal accountability does not. And AI does not face consequences. The people behind the contract do. That has always been true, and for now, it remains true regardless of what drafted the document.