top of page

Who is who? And who owns who and what? Artificial intelligence from the perspective of IP rights.

Artificial intelligence, in particular deep learning methods and techniques, is growing and developing rapidly. More tools and solutions, including AI-based systems, are becoming more and more sophisticated with – at least sometimes – some degree of autonomy. There is ongoing debate as to whether it is possible to create Artificial General Intelligence (AGI) and if yes – to what extent should it be allowed to enter legal relationships. Currently, however, there is a consensus that narrow AI should not have a (quasi) legal personality and should not be able to operate autonomously in the human environment.

At the same time AI systems can create ‘own’ content that is highly innovative, unique, and ‘personal’ (see Portrait by AI program sells for $432,000). Such content is based on datasets that were created in the past by humans and combined with AI-driven techniques and methods to create something new and original. One question here is whether AI, let’s say deep learning model can be treated as an innovator and therefore can be the owner of patents and the second – how to judge who may benefit from the outcome (also authors of the data that was used by AI?).

The answer to the first question is quite simple – in most cases, the bearer of the patent will be the owner of the AI system (or developer in certain cases). When it comes to the innovator that should be distinguished from the owner of the patent, the answer is not that straightforward. A few months ago the Australian Court has ruled out that the AI – in certain circumstances – can be treated as the inventor as it has some degree of autonomy. But if we go into details, we will quickly understand that the reason for such ruling was the ‘ineffective’ and outdated legal framework for intellectual property and not the mere fact that the AI is autonomous. In most cases (including in the United Kingdom) it is common not to grant AI the right to be an inventor. Nevertheless, the debate is still ongoing, and the future may bring new solutions – in the EU, however, there is no appetite to change the status quo. This question is, however, important as some legal regimes may not allow granting copyrights to the content not created by humans.

When it comes to the co-author's issue it is not that simple. As most lawyers say – the answer ‘depends’. Depends on the details of each case. The artificial intelligence system may use different data (that can be designs, paintings, or photography) that may or not be subject to intellectual property rights. If such data is not protected, e.g. is open source, the AI and its owner should not be worried. The case is, however, different if the AI is using data that is protected under relevant IP laws. This does not automatically mean that the outcome should be co-authored. The biggest challenge here is to assess to what extent such data was used by the AI to create new content. Is it a ‘copy’ of input data or it is just an inspiration? Is such input data a part of the outcome? How has the AI processed the data to create this new content? Are we able to explain such reasoning? If not, how can we assess to what extent that input data is a significant part of the work done by the AI?

No simple answers. Lawmakers and regulators are currently working on the appropriate regime for AI-specific IP law. In 2020 the European Parliament has presented its resolution on that matter. We are now waiting for the next steps. We will keep you posted!

Related Posts

See All

Document Management, is it a CEO's problem?

In a world of overflowing files and digital chaos, document management (DM) has emerged as a puzzle that begs for a solution. But should CEOs be the ones to crack it? Let's unravel the secrets of docu

The High Cost Of Commas.

The contract language is like software code: one small typo, and the whole thing falls apart. Running a successful business comes with its fair share of challenges, from managing finances and employee


bottom of page