Can computers learn and reason? If so, what are the limitations of the tasks that they can be given? These questions have been the subject of countless debate as far back as 1937, when Alan Turing published his work on computable numbers1. Many researchers have devoted themselves to developing methods that would allow computers to interact more easily with human beings and integrate processes used to learn from the situations encountered. Generally speaking, the aim was to have computers think and react like a human being would. In the early 1960s, Marvin Minsky, a noted MIT researcher, outlined what he regarded as the steps along the path to artificial intelligence2. The power of the latest computers and the capacity to store phenomenal amounts of information now allow for artificial intelligence to be integrated in business and daily life, using processes known as “machine learning”, “data mining” or “deep learning”, the last of which has undergone rapid development in recent years3.
The use of artificial intelligence in business raises many legal issues that are of crucial importance when companies enter into contracts respecting the sale or purchase of artificial intelligence products and services. From a contractual perspective, it is important to properly frame the obligations and expectations of each party.
For suppliers of artificial intelligence products, a major issue is their liability in the event of product malfunctions. For example, could the designers of an artificial intelligence system used as an aid in making medical decisions be held liable, directly or indirectly, for a medical mistake resulting from erroneous information or suggestions given by the system? It may be appropriate to ensure that such contracts expressly require that the professionals using such systems maintain control over the results, regardless of the context in which the system is operating, be it medical, engineering or business management.
In return, companies wishing to use such products must clearly frame their targeted objectives. This includes not only a stated performance objective for the artificial intelligence system, but also a definition of what would constitute product failure and the legal consequences thereof. For example, in a contract for the use of artificial intelligence in production management, is the objective to improve performance or reduce specific problems? And what happens if the desired results are not achieved?
Another major issue is the intellectual property of the data integrated and generated by a particular artificial intelligence product. Many artificial intelligence systems require the use of a large volume of the company’s data for such systems to acquire the necessary learning “experience”. However, who owns that data and who owns the results what the artificial intelligence system has learned? For example, for an artificial intelligence system to become effective, a company would have to supply an enormous quantity of data and invest considerable human and financial resources to guide its learning. Does the supplier of the artificial intelligence system acquire any rights to such data? Can it use what its artificial intelligence system learned in one firm to benefit its other clients? In extreme cases, this would mean that the experience acquired by a system in a particular company would benefit its competitors.
Where the artificial intelligence system is used in applications targeting consumers or company employees, the issues related to confidentiality of the data used by the artificial intelligence system and protection of the privacy of such persons should not be overlooked.
The above are some of the contractual issues that must be considered and addressed to prevent problems from arising.