22 Sep 2019 AI and IP: No man’s landscape?
Technology is evolving at an incredible pace. An ordinary smartphone has thousands of time the power of calculation of the computer that allowed the Apollo 11 to land on the moon. We are living the edge of a new era and, among other things, like quantum computer and gene editing, the Artificial Intelligence (“AI”) is going to have an enormous impact on everyday life.
Such technologies challenge the foundation of many legal aspects. One of the most interesting is the relationship between AI and Intellectual Property (“IP”). This newsletter will delve into aspects of such relationship.
Disclaimer: Readers of this newsletter should note that at the date of circulation of this newsletter, the laws and regulations to which this newsletter makes reference to may be subject to further amendments which will not be reflected in this newsletter.
What is AI?
It is quite complicated to define what is AI and, unfortunately, there is no definitive answer.
We should first differentiate between a “strong” AI and a “weak” AI. The first one is a machine capable to do all the things that humans can and in complete autonomy (in theoretical terms: it can fully ignore the limits that the famous computer scientist Alain Turing discovered in his works, among others, a software cannot code its own algorithms). The second is a very sophisticated software that can, with a certain degree of autonomy, complete one or more specific task (painting, evaluating fines, writing legal newsletters and so forth).
Many mathematicians and computer scientists believe that we will never see a strong AI, either because it is impossible to build or too complicated. They claim that we can only see a weak AI.
For the purpose of this newsletter, we will make reference to AI which is capable to develop IPs.
What are IPs?
It is easier to define what an IP is. It can be defined as a property that is related to the activities of an intellect. The typical examples are trademarks, designs, patents and copyrights.
Every one of them has its own peculiarities, but, generally speaking, they are all the result of a creative process. Most of them tend to be registered and disclosed to the public, but there is a certain protection also for unregistered IPs or trade secret.
The AI meets the IP
There are different ways in which IP and AI meet and this is already available.
In the artistic field, it has already been developed, for example, AI that writes pop songs has already been established (an app like that, called Alysia, can be downloaded).
Moreover, in the scientific field, there are many applications right now which make use of AI and a typical example is in Austria where AI is deployed to look for tumours in x-ray imagines, Also, in Denmark AI is used to analyse the voices of those calling the emergency services for patterns of cardiac conditions.
Furthermore, AI has also become part of our everyday life. An example of this is the new launch of the Google assistant which is able to perform simple tasks (booking, managerial organisation and so forth).
The future will bring more and more of these.
Is AI covered by IP?
An interesting issue is that AI itself could be covered by IP rights, be it patent or copyright.
However, this appertains to the more general issue dealing with the question of whether a software is patentable or not.
The current position of the EU is to allow the patentability of a software only under certain stringent conditions.
In order to be patented, a software must provide a new and non-obvious technical solution. This is often very hard to achieve or demonstrate, especially for small businesses, while corporation can invest more money in patent attorneys.
As such, a start-up creating AIs should carefully evaluate whether there is a strong basis for patenting AI itself.
On the bright side, AI would always be protected by copyright in the EU.
Is the development of AI in IP really the next big issue?
While the issue poses some problems, it is not utterly true that it is completely new.
It must be distinguished whether AI is helping the development of IP or whether it is developing it on its own.
In the first case, it is not so different from the situation of a coder, who is an employee of a company, writing a software. That IP will not belong to him or her, but to the company that has engaged and hired the coder.
Moreover, to a certain extent, AI devolving, for example, a database, is already covered by the so-called Database Directive, which allows IP rights on machine-created database.
In such case, a good old-fashioned lawyer-written contract, which clearly attribute the rights, or a license would be enough.
On the other hand, the situation of AI developing on IP by itself is less clear.
There are two extremely different approaches: On the one hand, it could be argued that a machine is not a natural person and therefore it is not capable to creating something original and new, which entails that IP created by such machine is not even IP, or, from the completely opposite side, there are authors who have proposed to apply the old Ancient Roman rules for slaves (and, therefore the copyright belongs to the AI proprietor). Moreover, there are also some authors that believe it is time to create some form of “electronic person”. Anyhow, such proposals have been rejected by the EU.
It is not possible to solve this issue abstractly, but it must be addressed evaluating what the AI is concretely doing.
What about liability?
If AI has violated IP rights for the creation of IP and AI was just helping the human, the ordinary rules for liability should be applied (a machine does not relieve a human from check and intervention).
In the case of an autonomous AI, it is not clear how to address the issue.
It mainly depends on the approach: if one believes that the old rules for slaves apply, then the owner is reliable. If it is not protectable, there is no liability. If it owns some form of “personality”, the AI itself will be liable.
More guidelines are expected to be issued on such topic and the case law is expected to provide further assistance.
Could the Blockchain technology help?
In this context, AI based on blockchain or any distributed ledger technology would not be able to solve any of the mentioned issues.
If AI itself is based on blockchain, the problems would not be solved. The fact that there is a public or private ledger would not add any legal value, in relation to IP, and it would probably complicate some other aspects (jurisdiction and applicable law across multiple countries and so forth).
If the AI generates blockchain algorithms, the already mentioned problems stand.
Technology is changing and shaping the world. New issues can be challenging, but not necessary able to stop business.
The relationship between IP and AI is already relevant and it is foreseen to become more important.
However, even if it seems something new, the law already provides the necessary flexibility to deal with the problem.
Data protection, technology and corporate matters.