Burst or Bust! The AI Contract Bubble

5 mins

Why AI contracts are doomed and their temporary success points to a failure in contract law.


MANY FUTURISTS IN THE LEGAL INDUSTRY speculate about the potential for artificial intelligence (AI) and machine learning in law. While it’s hard to predict their impact on law, it is reasonable to predict that artificial intelligence will find a place where there is no real intelligence involved. Contract law is not such a place, because a contract essentially requires human intelligence to make.

Yet some of the main applications of AI to law so far have been to draft and review contracts. The table below shows some examples.

Legal technologies that use AI and their applications to contract law

Application Legaltech or Company that Produces It
Document proofing Contract Companion (Litera Microsystems), Drafting Assistant (WestlawNext), Word Rake
Technology assisted due diligence eBrevia, Kira Systems, Luminance
Document automation Contract Express (Thomson Reuters), HotDocs
Contract analysis KM Standards, LawGeex, RAVN (iManage)

See Andrea Alliston’s article “Seven Legal Technologies Every Lawyer Should Know About”, republished by Canadian Corporate Counsel, Canada Business Law, May 2018.

Beats the Purpose!

Although many big law firms in Toronto are buying into a future of AI in contracts, this is a speculative bubble that has to burst, since AI cannot and should not replace contract writing or review. We don’t say this as luddites opposed to AI in law, but because these technologies defeat the purpose of the law. Indeed, any successful application of AI to contracts proves the failure of contract law, whose purpose is to establish a ‘meeting of minds’ of the contracting parties.

A contract is evidence that the contracting parties are ad idem, meaning there is a ‘meeting of minds.’ It is a misconception that the physical document called the contract is the actual agreement between the contracting parties. A contract is not an agreement, but evidence of an agreement. What makes a contract a contract is the actual agreement of the contracting parties. A contract, therefore, is not valid if the contracting parties are not ad idem. Justice De Sa simply explains the law at para. 37 in Murphy-Larade v. Inpanayagam, 2019 ONSC 715 (CanLII):

In order to form a contract, the parties must be of one mind as to the essential terms of the contract. To determine whether the parties reached a meeting of the minds, or consensus ad idem, the court applies an objective test. The court considers whether a reasonable person, apprised of all the circumstances, would believe the parties had reached an agreement. J.M.B. Cattle Corp. v. 2144032 Ontario Inc., 2015 ONSC 7372 (CanLII).

Thus, while AI may seem to be the ultimate contractor with its ability to quickly draft and review contracts, the ultimate contractors are and can only be the contracting parties. If the contracting parties are not ad idem, the physical document called a ‘contract’ is not a contract, it’s just a piece of paper that doesn’t mean or prove anything.

UK Prime Minister Neville Chamberlain brandishing Adolf Hitler’s pledge never to go to war again in 1938

Munich Pact

In fact, if a ‘contract’ is written in language the contracting parties didn’t read or write, or that they don’t even remember or understand, then instead of proving there is a ‘contract’ it will prove there is none! Thus, the success of AI in contracts reflects the failure of the law at many levels, including the following two.

First Failure: Searle’s Chinese Room

The law fails insomuch as the need or use for AI review of contracts reflects the fact that lawyers don’t read them, but only look for a density or connectivity of words that establish a presumable legal point (which is all AI programs can do).

This point is worth elaborating. Philosopher John Searle presents a famous thought experiment in his paper, “Minds, Brains, and Programs,” published in Behavioral and Brain Sciences in 1980, known as the Chinese Room. The Chinese Room basically imagines a computer that translates Chinese into other languages based on a set of rules. Most people would agree such a computer doesn’t understand Chinese. It just follows a set of rules. Proof is that a human following the same instructions couldn’t be said to understand Chinese.

Searle’s Chinese Room (Source: Wikicomms)

Image result for searle's chinese room argument

Similarly, understanding is of the essence of a contract. The contracting parties must have a mutual understanding of the terms of the contract, as well as any third parties adjudicating a dispute concerning it. If there is a legitimate disagreement concerning the meaning of the terms, then the contract will be more difficult or impossible to enforce.

Second Failure: It’s the Agreement, Stupid!

The next failure of the law is that AI contract review shouldn’t replace human contract review, since part of the point of a contract is for the contracting parties to review it. Contract review by a lawyer makes sure the contract is legal and that a third party reading the contract has the same understanding as the contracting parties.

With advanced semantic algorithms, AI review could help to ensure that a contract is legal. However, this would be difficult to apply to natural language contracts, given how natural languages (including much legal jargon) are quite random and evolve pretty randomly. Two people don’t use the same words the same way and attach different meanings, thoughts and emotions to them based on their experience.

This is why writing a contract is a fine art that seeks to express in common language the mutual understanding of the contracting parties. It’s also why review of a contract by a lawyer is important to make sure their understanding of the contract is the same as the contracting parties’, given the randomness of natural languages. Careful writing and review of a contract is the best way to test (and attest) its clarity and mutual understanding. For instance, if fifty random and independent people have the same understanding of a contract, then the contract is clear, which helps to establish a ‘meeting of minds.’ (To put it crassly, it’s unlikely that ‘everyone gets it’ except the two people who wrote it.)


While AI may seem to be the ultimate contractor, this is an illusion because the whole point of a contract is to prove a ‘meeting of minds,’ which means only the contracting parties can be the ultimate contractors. The illusion that AI can ‘do the work’ for the contracting parties arises because people often confuse the document called a contract with the actual contract. The document called a contract is evidence of a contract and the actual contract is the agreement. A ‘contract’ isn’t evidence if the contractors haven’t read or understood it or if it was written and reviewed by a computer like the computer in Searle’s Chinese Room. In fact, a contract that is written and reviewed by a computer proves there is no contract!

Since it defeats the purpose of the law, the AI contract bubble has to burst. Or, it has to bust, because it’s against the law.

A two-line note scribbled on a piece of toilet paper is more of a legal contract than a word-perfect 50-page automated document with an “I agree” checkbox you have to check to move on to the next screen anyway.

A further lesson that can be gleaned from this is that the law doesn’t always have to change to adopt some new technology, nor is technology always necessary to solve a legal problem. Just because AI can do some interesting things with documents, like draft and review them, doesn’t mean the law must adopt this technology if it has no legal use or defeats the legal purpose. Nor does technology always point to a solution or progress. Sometimes, it can point to failure or the way back. For instance, AI contract drafting and review highlights a fundamental failure of contract law and points back to a more basic methodology. Indeed, in an age where no one reads contracts and simply checks the ‘I agree’ box to move on, the law must move in the opposite direction to favour ‘toilet paper’ contracts over the best ‘Google contract.’

Leave a Comment

This site uses Akismet to reduce spam. Learn how your comment data is processed.