Smart legal contracts under English law - Part 2: Formation and interpretation

The Law Commission has concluded its investigation in to how the current law of contract is able to apply to smart legal contracts. Our previous post provided an introduction to smart legal contracts and the big picture challenges that English contract law will need to address. In this post we consider two fundamental contractual questions addressed by the Law Commission report:

1. Do smart legal contracts satisfy the requirements of English law, in order to be considered valid?
2. In the event of a dispute, how might the English courts interpret the terms of a smart legal contract, particularly where the obligations contained therein are not written in English or another natural language?

Formation

An English law contract requires:

1. agreement;
2. consideration;
3. certainty and completeness;
4. intention to create legal relations; and
5. compliance with the relevant formalities.

1. Agreement

An offer to be bound on specified terms and the consequent acceptance of such terms is unlikely to present any issues for a natural language smart legal contract. The court would consider the words and conduct of the parties, as it would for any other contract.

The Law Commission considered whether it was ever possible for the exchange of code (without natural language communications) to amount to offer and acceptance, noting that such offer and acceptance may be undertaken by computer systems (rather than humans).

The Law Commission concluded that it is a question of fact as to whether parties have “’held out’ their computer programs for the purpose of reaching an agreement”.

The report noted knowledge of the identity of a counterparty is not a requirement of English law, and that the pseudonymous nature of certain DLT systems does not preclude agreements being made in such circumstances.

2. Consideration

The nature of smart legal contracts (both in natural language and code) means that it is unlikely the parties and/or courts will struggle to identify any consideration. The case of deeds, where consideration is not required, brings separate challenges, which are discussed below.

3. Certainty and completeness

Natural language smart legal contracts will need to be examined in the normal way, to consider whether there is certainty and completeness.

Where an obligation is recorded solely in code, the report considered operational code would “provide a strong indication that the agreement is certain and complete, because only complete and syntactically correct code will be executed by a computer.” Where the output of the code is not sufficient to deduce the legal agreement, the court would need to apply the ordinary rules of interpretation. The risk is that, where the ordinary rules of interpretation cannot resolve the issues of certainty and completeness, the contract may be deemed void.

In situations where a smart legal contract contains both natural language and coded terms, there may be conflicting obligations. Such conflicts would need to be resolved using the same rules of interpretation as discussed above.

4. Intention to create legal relations

The intention to create legal relations is a requirement of English law, and is assumed to be present where the parties make an express agreement in a commercial context.

In the context of a natural language smart legal contract this is unlikely to provide any new legal issues. However, where an agreement is to be inferred from parties’ conduct as the result of an interaction on a distributed ledger (as opposed to an express agreement concluded on a distributed ledger) it may be necessary for the parties to show that they intended to create legal relations.

Again, the report concluded that such a matter would be a question of fact, with the English courts having found binding “agreements based solely on the parties’ conduct where doing so is necessary to give “business reality” to their transaction, or where, in the circumstances, the parties would have expected enforceable obligations to exist.”

An influential school of thought, embodied in the phrase ‘code is law’, argues that enforcement by the courts should be completely excluded in favour of automated DLT procedures. The Law Commission report takes the position that this could be achieved by an express statement, perhaps in a framework agreement governing all agreements on a particular DLT, denying an intention to create legal relations:

“Some parties may be content to rely on the operation of the code and the relevant DLT protocol to do this and may not want to have recourse to the courts if something goes wrong. Such parties may expressly deny an intention to create legal relations, for example, by including a provision to this effect in the natural language component of their agreement. The courts have given effect to clauses which expressly deny an intention to create legal relations.”^[1]

5. Compliance with the relevant formalities

English law requires that certain contracts must be made in writing and signed (e.g. a guarantee).

In writing

Writing is defined as including “representing and reproducing words in a visible form”^[2]. Natural language smart legal contracts would comply with this requirement.

Before considering whether code-only smart legal contracts comply with this requirement, we need to consider that there are different types of code. By way of a gross oversimplification, source code is programming language in which human programmers can write instructions and high-level computer operations take place. When this source code is run, it is has to be translated (by a compiler or interpreter) into 1s and 0s for the computer’s various processing units to action. These 1s and 0s are a machine code and while it is possible for humans to read machine code, the process would be impractically time consuming.

The Law Commission considered that it was a much more difficult question for contractual obligations recorded only in code. The report concluded that the key issue was whether the code was in a form that could be read by a person, and that while source code is capable of being read by humans (being written in words and numbers), codes that reside in “machine code or a lower level of code than source code” may not come within the definition of writing.

While a distinction based on human readability is a helpful and useful guide, it is important to remember that all digital representations of information – from natural language, through source code and down to the lowest levels of code – are a string of 1s and 0s, which may be rendered readable to a greater or lesser extent. The Law Commission’s proposed test of human readability is an insightful solution, but will still require resolution of difficult questions of degree by the courts.

Any test based on human readability would also face complications where smart legal contracts are based on the output of AI deep learning – where computers program themselves in ways that humans cannot understand and may be fundamentally incapable of understanding and where the reasoning cannot be reproduced in human readable form.

Signed

The Law Commission did not foresee any issues with the requirement for a signature, with digital signatures ordinarily satisfying any statutory requirement for a signature. However, it was noted that specific contractual arrangements, legislation or case law may specify that a digital signature is not sufficient.

Deeds

In order to create a deed, a signature must be witnessed and attested. In the case of natural language smart legal contracts, this could be achieved in the ordinary way. The report expressly doubted that the current English contract law “supports the creation of deeds which are wholly or partly defined by code”, as the witness must be physically present upon signing. While the Law Commission considered that a witness could be physically present upon the digital signature of the code, they noted the requirement in Mercury^[3] that the signature and attestation must be on the same physical document.

Interpretation

Natural language smart legal contracts are unlikely to present novel issues of interpretation. The natural language contract will record the obligations, and the ordinary rules of interpretation would apply to determine whether the code reflects the agreed terms.

By contrast, coded instructions are not interpreted, but rather executed as instructions by the computer. The Law Commission considered whether this meant that code-only obligations were open to contractual interpretation, and concluded that they should be, as:

“We do not think that the code simply means what the code does when it is executed, or that it has no meaning, and only an effect. In our view, there can be a divergence between what the code “means”, and what it does when it is executed, which entails a distinction between meaning and effect.

Two potential methods of interpretation were discussed:

1. How would the coded term be understood by a functioning computer?

This methodology would consider only the effect of the code, i.e. the exact opposite of the Law Commission’s above conclusion. While this method is a possibility, it was not recommended by the report.

2. What would a reasonable coder understand the code to mean?

The report dismissed the man on the Clapham omnibus out of hand, preferring the ‘reasonable coder’ test.

The reasonable coder would assist the court by explaining “the effect of certain combinations of words, and give their reasoned opinion as to what the code appeared to instruct the computer to do.”

The Law Commission considered that the reasonable coder approach would result in an approach that is more closely aligned to the existing approach to natural language contractual interpretation. However, unlike traditional contracts - which cannot be executed to see how they might function – an instinctive first step for the reasonable coder would be to execute the code to see how it runs, which in turn is likely to affect the reasonable coder’s interpretation of code, so the distinction suggested by the Law Commission may be difficult to maintain in practice.

Key Takeaways

The formation of smart legal contracts under English contract law is not without challenges. However, with the exception of deeds – which present their own potential problems – the challenges appear to be sufficiently manageable that common law will be able to adapt to address them.

The Law Commission’s preferred test for the interpretation of code-only obligations is designed with the aim of aligning such interpretation with the approach taken to traditional contracts. However, the authors consider that the reasonable coder test has significant limitations, in that it overlooks the reasonable coder’s most natural first step of running the code, and secondly that it is not futureproofed for AI-generated code.

The final post in this series considers a range of dispute-focussed challenges - remedies, consumer contracts, jurisdiction, and governing law - that smart legal contracts may throw up.