In 1996, Nick Szabo coined the term “smart contract” in a revolutionary article on the introduction of digital technology in the realm of contracts. In that article, Nick Szabo stated: “[n]ew institutions, and new ways to formalize the relationships that make up these institutions, are now made possible by the digital revolution. I call these new contracts "smart" because they are far more functional than their inanimate paper-based ancestors. No use of artificial intelligence is implied. A smart contract is a set of promises, specified in digital form, including protocols within which the parties perform on these promises.”[1]

Since 1996, the emergence of blockchain technology has allowed the creation of a new form of legal ledgers which are run in a decentralised manner.

There are various classifications of smart contracts and smart legal contracts depending on whether they are fully or partially automated and recorded in a blockchain system or not.[2]  In this article, we will use the expression “smart contract” in its simplest and widest form.

We define smart contracts as computer instructions that represent the intention of the parties to create an obligation, make a payment, purchase an asset or service or trigger an event that has a legal consequence.

The primordial example of a smart contract is a vending machine. By inserting a coin in a vending machine, the person who inserted the coin purchases a snack or a beverage from the vending machine owner or operator. Automated ticketing machines are also old automated or self-executing contracts, whereby a person acquires the right to use a transport service, enter a movie theatre etc.

More recently, online contracts with Amazon and the like are also forms of automated purchase agreements that fall into the concept of “smart contract” because they allow clients to purchase goods and services by giving automated instructions through a machine.

In any of the above examples, there is an underlying “natural language” contract, that is, a language used by people and not coded instructions given to a computer. When we enter an online contract, we are often required to accept a standard contract, sometimes in a foreign language that we may not fully understand. This raises issues of the validity of those rules concerning general and standard clauses. We will not be concerned with these implications in this article; we will focus on the issue raised by “software coded contracts”, which is the widest possible definition of smart contracts.

“Software coded clauses” or “smart clauses” are computer instructions using software language that may constitute the performance of an obligation or trigger the verification of contract conditions. Software coded obligations are self-executing because, in most cases, they do not need human intervention. Software coded contract provisions are now used in all online contracts because every service or product purchased online triggers legal consequences, such as the obligation to pay and the obligation to deliver a service or good.

The entry and performance of online contracts have occurred without disrupting legal systems. Generally, around the world, legislations accept the validity of online contracts. Disputes have been resolved in a fairly satisfactory manner because suppliers that value their clients are ready to solve complaints in a friendly way. Customers dissatisfied with suppliers simply cease to purchase from those suppliers. The rules of supply and demand that govern mature and fair markets take the stress out of the system. However, some terms and conditions and the conduct of many online suppliers may not be legal on many occasions.

The buzz about smart contracts goes much deeper than simple online B2C online agreements, where software coded information is translated into service and purchase orders and money transfer instructions.

Smart clauses coded in self-executing software language can include complex payment terms and formulas, trigger conditions, default clauses and security enforcement provisions. Smart contracts can be used in all types of contracts, including very complex B2B contracts such as credit agreements, mergers and acquisitions, issuance, acquisition and disposal of securities, IPOs, financial derivatives, futures markets, restructuring agreements, construction contracts etc.

Also, matters of technical nature can be linked to the contract clauses through smart clauses, that is, technical parameters stated in legal contracts involving the operation of telecommunication networks, power grids, power production, software requirements etc. can be linked to the legal language of contracts and trigger contract consequences. Presently, many of those technical issues are left in the contract’s fringes and put into more vague and elusive contract terms, such as “best efforts” and reasonableness judgements, or left to mediation or litigation when they become unsolvable by the people dealing with the day-to-day operation of the contract.

There are five basic practical rules for ensuring that smart clauses are enforceable and do not raise even worse issues than natural language clauses.

First, smart clauses should be translated into a natural language. This means that every smart clause should have an equivalent natural language clause. The natural language clause should be as objective as the smart clause and not include open concepts that cannot be translated into the coded clause.

The need for this rule does not result from the Law but serves a practical need: clauses should be understandable by people without very deep knowledge of the law and the technical aspects of the contract. Judges and business decision-makers should be able to understand the key points of the contract.

Suppose the main clauses of the contract, like the performance of the contract, are dependent on a coded clause (with no natural language equivalent). In that case, it won’t be easy to understand why the parties chose that particular solution instead of another. Sure, there are contracts with heavy and complex technical issues, be that the legal formulation of the clause or the commercial and technical aspects. But, in most cases, the economic, commercial and technical terms have a “natural language” formulation inscribed in the agreements and understood by the persons present at the negotiation table, even when they include complex technical schedules.

Second, the content of smart clauses should be open and auditable. This means that the acceptance of the code should be made by experts of each of the parties. This requirement aims to give an equal footing to the parties. Smart clauses must be understood and controlled by the parties.

There should not be one party controlling the code and the consequences of the instruction generated by it. In present natural language contracts where the legal and commercial technicalities can be generally understood by the parties, each party should have its own legal advisors. In coded language contracts, an asymmetry in knowledge can be more damaging than not having legal counsel. Expert advice and verification procedures are necessary.

Third, smart clauses should be protected. Integrity is a key element of any contract. In natural language contracts, one of the parties cannot change the wording of clauses. This is ensured by putting the contract in writing and other formal requirements imposed by law or agreed by the parties. Tampering with the words of a contract means falsifying the contract content. The same rules apply to smart clauses and smart contracts.

However, because smart clauses are generally self-executing, the consequences of possible tampering with the contract software code are more direct and can unlock a chain of events that may be impossible to stop. For this reason, the integrity of smart clauses should be ensured.

Blockchain systems are an acceptable form of ensuring the integrity of contracts because the blocks of a blockchain cannot be changed without the agreement of the nodes of the system (all or a significant number of participants in the system depending on the type of consensus algorithm that is used). This ensures the integrity of the contract in a manner that is as efficient, if not more, than current government-sponsored or other centralised ledger systems. However, blockchain is not the only way to ensure integrity. The parties may appoint an independent third party to store and protect the code or even monitor its application.

Four, smart clauses triggering events that require human intervention should not be left to the parties. While many smart clauses rely on machine self-execution mechanisms, there are cases where human intervention is necessary to fill gaps or interpret the data. The parties should not take this decision.

For instance, if the contract stipulates that one of the parties must make a payment to the other if the temperature reaches 45 degrees and two official computer records state different temperatures, one 44.9 degrees and another 45 degrees, a decision will have to be made as to whether the payment condition was met or not.

This example underpins the fact that minor discrepancies in digital records or the non-existence of an independent digital record may require human intervention to verify or certify smart contracts trigger conditions. When that is the case, the person making that decision should be independent of either party.

In Ethereum, the contract parties may appoint persons, named “oracles”, to take decisions that will trigger or not trigger smart contract events. This is an adequate solution for smart contracts based on Ethereum. For smart contracts outside a blockchain, the parties may use independent entities and grant them the power to fill in the missing data or resolve digital records inconsistencies.

Five, contracts with embedded smart clauses should include effective and fast dispute resolution mechanisms. Smart contracts should contain dispute resolution mechanisms because smart clauses can add a layer of complexity for which courts are unprepared.

In existing b2c online contracts, few disputes are resolved in courts because their value is low. The consumer often simply refrains from purchasing from the vendor that failed to deliver. This is not the ideal way of resolving a dispute, and a more effective transnational redress system must be created.

In more valuable contracts, disputes can be taken to courts, but the time for resolution may be too long to remedy the harm suffered by one of the parties. Parties will trust more in contracts that contain safe and expedient means of resolution. These may include mediation and arbitration mechanisms that allow for taking over control over the code, stopping wrongful use of the code, correcting an improper use of the code and repairing or correcting the code that proved not to attain the parties’ goals.

Smart contracts are one of the greatest inventions of the turn of the century. They improve efficiency increase speed and quality of delivery.

Embedding smart clauses in natural language contracts and creating adequate dispute resolution systems backed by robust control and verification mechanisms are the first steps to make your contracts smart.


A shorter version of this article was published by Público, a Portuguese newspaper, on January 18. To read, follow this link


The idea that robots and machines might one day replace people fascinates and scares many people. Literature, movies and cartoons depict the image of robots that take over control over humankind in a not-so-distant future. The Matrix saga, 2001: A Space Odyssey, The Jetsons, among many others have populated the imagination of children and adults for generations. Until the beginning of the XXI century, the subject of artificial intelligence (AI) was confined to the academic world, science fiction, and a few industries. At the end of the last century, Goggle brought AI to everyone’s lives in a palpable way. Research at our fingertips replaced long hours in libraries and bookshops.

Today, AI is everywhere, in social networks, in the advertising industry, computer research, development of cars and other machines among so many other things. Computer software is powered by AI engines. Smartphones, computers and laptops have embedded AI. When you lift a smartphone, unlock your phone using face ID or use speech recognition, you are using AI-powered software.

But the AI revolution seems not to have reached the business of law. Of course, lawyers and law firms now use AI in their electronic devices and in computer software, some of which is designed specifically for lawyers, such as document management systems and billing software. But AI has not disrupted the business of law as it has done in so many other businesses.

The changes to the business of law brought about by legal AI technology, the concerns and scepticism of lawyers call for a discussion about the role of legal AI in the future and if some or all of the lawyers’ tasks will be replaced by legal machines.

First, let us define AI. According to the Encyclopaedia Britannica “artificial intelligence (AI) is the ability of a digital computer or computer-controlled robot to perform tasks commonly associated with intelligent beings”. AI is the ability to simulate the logic of an algorithm by a machine. An algorithm is a finite sequence of defined instructions used to perform a computation. Algorithms are used as specifications for performing calculations, data processing, automated reasoning, automated decision-making and other tasks. AI gives machines the ability to perceive a given environment and to take actions to achieve the goals set by the machine’s program.

To some lawyers, this is impossible, at least for the more complex legal matters, because the business of law, whatever its shape and form, lives around “words”. From the dawn of history, enacting and enforcing laws has revolved around the use of “words”. Rhetoric and grammar have always been at the core of all legal professions. Words may be ambiguous and have multiple meanings depending on the context and order in a sentence. The interpretation of words appears to be a purely human activity.

However, due to its “prescriptive” nature, the Law implies simple deductive reasoning processes, which makes it ideal for AI “coding”. Putting it simply, the Law is not immune to AI; on the contrary, the Law is an ideal field for AI.

The good news for today’s lawyers is that we are far from the day when AI will take over their business. For that to happen, we must “code legal thought”.

To code legal thought one needs to create the processes that will allow machines to interpret laws, contracts and court decisions, which seems still far on the horizon because “words” and “sentences”, the matter of law, have many meanings, many times elusive, and open to manipulation by legislators, the politicians that make the laws, lawyers and even judges, who are no more than humans and, therefore, have sentiments, opinions and convictions to which they hold strong. Right and wrong are not clean-cut, are not yes and no, are not a series of 0s and 1s.

Yet, legal thought can and will be software coded in the not-so-distant future.

This is how it will happen. Firstly, AI will replace lawyers and other legal practitioners in the review of documents. Document review systems use machine learning technologies and pattern recognition technology to identify key contract concepts, tag clauses, court decision patterns, flag discrepancies and other patterns in the application and interpretation of laws and contracts, etc.

Lawyers are still needed to interpret the data coming out of computer systems and attribute meanings to those discrepancies and patterns. The machine will learn from the human lawyers’ interpretation, build new models, discern potential risks, etc.

Existing computer programs, like Luminance or Kira, already successfully help lawyers to conduct due diligences and provide advice to clients.

The second stage of legal AI will affect how courts prepare their rulings. AI-powered research systems already give judges and lawyers access to legal precedents concerning the specific subject matter of the case. In the future, AI will enable judges to narrow down the key elements of their decisions and offer them a roadmap for the decision-making process. Let us take a simple court decision such as determining if the court has jurisdiction over a particular matter that is put to it. Legislations around the world have a clearly defined set of rules for determining courts’ jurisdiction, which can be coded in computer language, that is, in an algorithm.

Some software systems, such as Lexis+, offer court decisions analytics and help lawyers to assess the likelihood of success of cases based on past decisions.

Future legal algorithms will help legal professionals to determine if matters follow into one legal category or another and how the law will be applied in specific cases. This will be one step away from determining the application of legal rules. Again, taking a simple example, if someone goes into another person’s property and it was not entitled to do so by law, such action falls under the legal concept of trespassing. These tasks will be done with an increasing degree of complexity, eliminating false positives, the application of conflicting rules and rights, the existence of exculpatory reasons to dismiss the application of a certain rule and the choice for another prevailing rule.

Many lawyers will argue that doing law has some specific features, such as the interference of sentiments and beliefs, which make it a non-computable task. AI can probe into statistics and the cold walls of legal statutes - they will say - but not into the hearts and minds of real people.

This is the wrong approach to legal AI. There are two areas where legal AI will struggle to master: an upper layer of present laws with their cultural, sentimental and political veneer that cover the law’s core rules. This veneer will be cleaned up by more powerful algorithms and the rationality of legal algorithms, uncovering a simpler and, therefore, fairer set of rules free of many of their present inconsistencies and conflicts.

As far as lawyers are concerned, legal technology and, in particular, AI systems will change the way they work but will require all of their ingenuity to interpret the complex legal issues that are needed to power legal AI. Lawyers and law researchers will have to break down the questions embodied in existing laws, contracts and legal precedents to map the AI systems of the future. Lawyers will also have to understand how to introduce queries into AI-powered systems and interpret the results brought to light by AI, formulate legal strategies and create legal documents, such as court briefings, contracts and opinions.

Legal AI will empower lawyers and clients. Legal AI will create faster and more efficient ways of completing all legal tasks. This is the case of knowledge management, documentation analysis, contract drafting, litigation review, preparation of court briefings. Legal AI will release lawyers from many repetitive and standardised tasks.

AI will disrupt the business of law. AI will lead to the disintermediation of legal services as it is now doing in other business sectors. In finance, for example, centralised exchanges will have to compete with the blockchain-enabled trustworthy peer to peer (P2P). The same will happen with many existing legal services. The traditional legal services will only be provided by lawyers when a trusted confirmation or reliable advice cannot be found in other sources.

Many lawyers view the disintermediation of other business sectors as unwelcome news, unaware that automated legal services have already taken a chunk of their old business. Automated services now offer many services that were exclusively provided by lawyers and paralegals a few years ago, such as invoice collections, automated contract drafting is now offered to business consumers and for low-value contracts. It is foreseeable that in the medium to long term, many legal services will be taken over by machines.

All legal professions will benefit from AI. Lawmakers will make better laws; judges will issue fairer judgements; lawyers will be able to perfect their court pleadings to a higher degree of efficiency and practicality. Many conflicts will be solved before being taken to courts because the odds, that is, the likelihood of success, will be against one of the parties in a clearer way, prompting that party for a settlement or simply to avoid litigation.

AI will not replace lawyers, but it will radically change the way lawyers provide services.

Legal thinking must remain in the sphere of lawyers because inside the core of all laws live values and values are non-computable. Values cannot be reduced to the mathematical formulations of algorithms. The creation and the application of law must, in the end, be made by humans and for humans.

Law is a science and a technique, but it is also an art and therefore cannot be reduced to mere algorithms. This is the limit of legal AI and the limit to any AI-powered technology.


The “Startup Visa” is a hosting program focused on foreign investors who intend to develop an enterprise or innovative project in Portugal. Such entrepreneurs may apply for a residence authorization/visa if certain requirements are met. 

This program establishes that only “certified incubators” can be responsible for supporting the development of new businesses, the provision of equipped spaces, and administrative and marketing assistance under an incubation agreement eligible candidates.

The certification process for incubators within the “StartUP Visa” program is now completed and the final list of “certified incubators” is now available at IAPMEI’s website


The “Startup Visa” is an hosting program focused on foreign investors who wish to develop an enterprise or an innovative project in Portugal. Such entrepreneurs may apply for a residence authorization/visa if:

  • they have a real and effective interest in developing in Portugal an entrepreneurial project, including but not limited to the creation of innovation-based companies;
  • the activity carried out through such project aims at the production of international and innovative goods and services;
  • the project has potential to create qualified employment (at least 5 jobs in 24 months); and
  • one or more “certified incubators” have shown interest in incubating the project.

A list of “certified incubators” will be made available to entrepreneurs on the IAPMEI’s website, during the month of February. These incubators will be responsible for validating the projects, for supporting the development of new businesses, for the provision of equipped spaces and for administrative and marketing assistance under an incubation agreement.

For each application, 5 residence authorizations/visas may be requested. The application can be submitted electronically in Portuguese or in English, to be examined by the IAPMEI.

The benefits granted to entrepreneurs under the “Startup Visa” shall be in force until the end of the incubation agreement.


IFLR 1000

"Excellent service, says a client on a refinancing transaction: I always trust on them when I have a legal issue in Portugal. [They are]...

Chambers and Partners

"Macedo Vitorino remains the go-to legal adviser for major telecoms players and specialises in the financing of public companies. Recent...

Chambers and Partners

"A compact but technically strong team, recognised for its corporate and M&A experience. Provides particular expertise in the TMT, energy...