Authored By: Barirah Shaikh
Middlesex University Dubai
Introduction
Traditionally, contract formation has relied on human actions and intentions grounded in the principles of offer, acceptance, consideration, and intention to create legal relations. English law recognises contracts made through electronic communications and automated systems. This means that the absence of direct human involvement does not make a contract invalid. However, modern artificial intelligence systems are different from traditional automated systems because they can negotiate terms, make decisions, and accept offers with limited human involvement. As a result, an AI system may enter into an agreement that its user did not specifically approve. This creates uncertainty about whether the agreement should be binding and, if so, who should be responsible for it. It also raises questions about the AI’s authority, the parties’ intention and the effect of mistakes made by the AI.
This article argues that AI does not need to have its own legal personality for a contract made through it to be legally binding. When a person or organisation chooses to use AI to negotiate contracts, the agreements made through that system are generally legally binding. However, when AI acts outside the authority given to it or makes a significant mistake, existing principles of contract law need to be considered.
The article first examines the rules governing electronic and automated contracts under English law. It then considers relevant case law and whether these rules are sufficient to deal with AI-generated agreements. Finally, it considers the UNCITRAL Model Law on Automated Contracting 2024 and explores how English law could respond to the growing use of AI in contract formation.
Contract Formation and Its Relation to Automated Systems
To understand whether an AI-generated contract can be legally binding, it is necessary to understand what makes an agreement a contract under English law. A contract is a legally enforceable agreement between two or more parties. It is generally formed when one party makes an offer, and the other accepts it, with consideration and an intention to create legal relations. In Carlill v Carbolic Smoke Ball Co, the court considered the role of offer and acceptance in forming a contract.[1] In Currie v Misa, consideration was explained as something of value given in return for a promise,[2] while Balfour v Balfour shows the importance of an intention to create legal relations.[3] Courts generally look at the words and actions of the parties objectively, rather than simply asking what either party privately intended. This approach can be seen in Smith v Hughes.[4] These requirements help determine whether the parties have reached a legally binding agreement. These same principles provide the starting point when a contract is made through AI. The issue is not whether the usual rules of contract formation apply, but how those rules should be applied when AI is involved in communicating, negotiating or accepting an agreement.
The use of technology in contract formation is not something completely new. In Thornton v Shoe Lane Parking Ltd,[5] the customer entered into a contract through an automatic parking machine without dealing directly with a person. The court accepted that a contract could be formed through the operation of the machine. The machine was not treated as a separate legal person. It was simply the means through which the contract was made. This shows that a contract does not become invalid simply because a machine is involved.
AI takes this principle a step further. Unlike a simple machine that usually follows fixed instructions, AI can analyse information and make decisions. For example, an AI system could negotiate a price, make an offer or accept an offer on behalf of a business. This raises an important question: if the AI makes the decision, should the business be legally bound by it?
The AI itself does not necessarily need to be treated as a legal person for the contract to be valid. If a business chooses to use AI to negotiate or enter into contracts, the AI can be viewed as a tool through which the business conducts its activities. The more important question is whether the AI was acting within the authority given to it by the business. This becomes particularly important where the AI makes a decision that the business did not expect or acts outside the limits set for it.
AI System and Its Authority in Contract Formation
There is a strong argument that a business should be bound when its AI acts within the authority given to it. English law already recognises that a person or company can be bound by the actions of an agent when that agent is acting within the authority given to them. This principle can reasonably be applied to AI. If a company chooses to give an AI system the authority to negotiate or enter into contracts, there is a strong basis for treating agreements made within that authority as binding on the company. The fact that the decision was made through AI rather than directly by a human should not make the agreement invalid.
This approach also gives the other party some certainty. If a customer is dealing with an AI system that has been authorised by a company, the customer should generally be able to rely on the agreement. Otherwise, a business could avoid responsibility by saying that it was the AI, and not a human employee, that made the decision.
The situation becomes more difficult when AI goes beyond the authority given to it. For example, a company may tell its AI that it can make purchases up to 50,000 euros, but the AI may make a purchase worth 100,000 euros. In this situation, the law would need to consider what the AI was allowed to do and what the other party knew about those limits. Existing rules on authority can help in these situations, although they were mainly developed for situations involving human agents.
Mistakes can also cause problems. An AI system may make a decision that the user did not expect or intend. If this happens and the mistake is serious enough to affect the contract, the existing rules on mistake may become relevant. However, it is still unclear how these rules will apply to AI systems that can make decisions on their own
This means that AI-generated contracts should not simply be treated as either always valid or always invalid. Instead, each situation should be considered on its own. If a business has given its AI permission to enter into contracts, there is a strong argument that the business should be bound by contracts made within that permission. If the AI goes beyond what it was allowed to do or makes a serious mistake, the situation may be different.
The main point is that AI does not need to become a legal person for a contract made through it to have legal consequences. The contract can still be between the business and the other party, with the AI simply being the tool used to make an agreement. This approach is also supported by the Law Commission, which has found that existing English law can generally accommodate contracts made through automated systems.[6] However, as AI becomes more independent in making decisions, existing legal rules may need to be applied to new situations that were not considered when those rules were developed.
The Role of UNCITRAL Model Law on Automated Contracting
The legal questions surrounding AI-generated contracts are not limited to England. As businesses increasingly use automated systems to make and manage contracts, other countries have also started considering how the law should deal with these transactions. One important development came in 2024, when the United Nations Commission on International Trade Law (UNCITRAL) adopted the Model Law on Automated Contracting.[7] The Model Law is important because it provides a legal framework for contracts made through automated systems, including contracts made by artificial intelligence.
One of the main ideas behind the Model Law is that a contract should not become invalid simply because a person was not directly involved at every stage of its formation. This is particularly relevant to AI, as an AI system may communicate with another party, make a decision or produce contractual terms without the user checking every step. However, this also raises an important question: what happens when an AI system produces a result that the person using it did not expect? The Model Law addresses this situation in Article 8, which deals with unexpected actions by automated systems.[8] Where an automated system takes an action that the person using it could not reasonably have expected, and the other party knew or should reasonably have known that the action was unexpected, that action may not be relied upon by the other party.[9] In simple terms, the Model Law does not automatically cancel a contract just because an AI system has made a mistake. Instead, it considers how unexpected the action was and whether the other party knew, or should have known, that something had gone wrong. This provides protection when AI produces a result that goes beyond what its user could reasonably have expected.
The way Article 8 deals with unexpected actions is important because it shows that the law can deal with problems caused by AI without treating AI as a separate legal person. The focus remains on the people or businesses involved in the contract and on what happened when the AI made the decision. In other words, the AI does not become a party to the contract simply because it was involved in making the agreement.
For English law, this suggests that there may not be a need to create a completely new set of laws for AI. The existing rules of contract law can continue to apply to AI-generated contracts, while approaches such as Article 8 of the UNCITRAL Model Law can help deal with situations where an AI system produces an unexpected result. Although Article 8 is not binding in England, it shows that such problems can be addressed without treating AI as a separate legal person. Instead, the focus can remain on the parties involved and on the circumstances in which the AI made the decision.
This becomes important as businesses give AI greater responsibility for negotiating and making contracts. If businesses are going to rely on AI in this way, they need to know when they will be responsible for the agreements it makes. At the same time, the other party needs to know whether an agreement made through AI can be relied upon. Therefore, the aim should be to allow businesses to benefit from AI while also providing clear rules about when an AI-generated contract will be legally binding.
Conclusion
The increasing use of AI in contract formation does not require contract law to treat AI as a person or as a new type of contracting party. The better view is that AI remains a means through which people and organisations communicate, negotiate and enter into agreements. The fact that a contractual decision is made by an AI system rather than directly by a human should not, by itself, prevent the agreement from having legal effect.
The real difficulty lies in determining when the actions of an AI system should be legally attributed to the person or organisation using it. Where a business deliberately gives an AI system authority to negotiate or conclude contracts, there is a strong basis for holding the business to agreements made within that authority. This is necessary not only to maintain responsibility on the part of businesses, but also to protect the reasonable expectations of those who contract with them. A party should not be able to avoid an agreement simply because the decision was made through an automated system.
However, there must be limits to this principle. AI systems may act beyond the authority given to them or produce outcomes that their users could not reasonably have anticipated. In such circumstances, the law must consider the nature of the authority given, the seriousness of the mistake and what the other party knew or ought reasonably to have known. This allows the law to distinguish between an ordinary contractual decision made through AI and a genuinely unexpected or unauthorised outcome.
The UNCITRAL Model Law on Automated Contracting demonstrates that these difficulties can be addressed without giving AI its own legal personality. Its approach to unexpected actions provides a useful direction for dealing with situations in which automated systems produce results that fall outside reasonable expectations.
Overall, the law should not focus on whether AI itself intended to enter into a contract. It should focus on the people and organisations that choose to place AI in the contracting process and on the circumstances in which the resulting agreement was made. Therefore, English law does not need to create a separate law for AI-generated contracts. What it needs is to apply its existing principles clearly and consistently as AI becomes more autonomous.
In a nutshell, the involvement of AI does not alter the fundamental principles of contract formation or require AI to be recognised as a legal person. What matters is whether the actions carried out through the AI can properly be attributed to the party using it. If English law can maintain this, then it can accommodate the increasing use of AI in contracting while preserving the certainty, fairness and accountability that are essential to a legally enforceable agreement.
Bibliography
Primary Sources
Cases
- Balfour v Balfour [1919] 2 KB 571 (CA)
- Carlill v Carbolic Smoke Ball Co [1893] 1 QB 256 (CA)
- Currie v Misa (1875) LR 10 Ex 153 (Exchequer Chamber)
- Smith v Hughes (1871) LR 6 QB 597
- Thornton v Shoe Lane Parking Ltd [1971] 2 QB 163 (CA)
Secondary Sources
Websites
- Law Commission, Smart Legal Contracts: Advice to Government (2021) <https://lawcom.gov.uk/project/smart-contracts/> accessed 6 September 2026
- UNCITRAL Model Law on Automated Contracting with Guide to Enactment (United Nations 2024) art 8 <https://uncitral.un.org/sites/default/files/2424674e-mlautomatedcontracting-rev1-ebook.pdf> accessed 6 September 2026
[1] Carlill v Carbolic Smoke Ball Co [1893] 1 QB 256 (CA).
[2] Currie v Misa (1875) LR 10 Ex 153 (Exchequer Chamber).
[3] Balfour v Balfour [1919] 2 KB 571 (CA).
[4] Smith v Hughes (1871) LR 6 QB 597.
[5] Thornton v Shoe Lane Parking Ltd [1971] 2 QB 163 (CA).
[6] Law Commission, Smart Legal Contracts: Advice to Government (2021) <https://lawcom.gov.uk/project/smart-contracts/> accessed 6 September 2026.
[7] UNCITRAL Model Law on Automated Contracting with Guide to Enactment (United Nations 2024) art 8 <https://uncitral.un.org/sites/default/files/2424674e-mlautomatedcontracting-rev1-ebook.pdf> accessed 6 September 2026.
[8] ibid.
[9] UNCITRAL Model Law on Automated Contracting (n 7) art 8.
