AI as an hybrid asset

Focus on AI gives you an overview on how technical, legal and business aspects relating to AI intertwine with each other and form a novel, hybrid type of an asset.

It contains:

  1. Walkthrough to the components and assets involved in building AI and a visualisation on their relation to forms of IP protection
  2. Copyright issues with regard to AI
  3. Applicability of hybrid intellectual property to AI
  4. Basic features needed to build truly Open Source AI
  5. Promoting openness to respect software freedom
event-2024-03-26-startups

1. Walkthrough to the components of AI and their relation to IP protection

To understand the technical, legal and business aspects relating to AI we should first take a look at the pipeline of machine learning. On a general level, the pipeline consists of actions of building the untrained model, getting the training data, iterative training the model, applying the trained model to data and using the output.

Each of these actions entails several steps. Each of the steps requires human-computer interaction and may be quite laborous.

To be able to understand, how these actions intertwine with forms of IP protection, we can divide the IP protection into a spectrum of items, some of which are protected by various forms of IP protection, some do not receive any form of protection. The requirements vary depending on the form of IP protection.

If we now look in more detail the components that are either fed into the machine learning process or are generated during it, we can start to identify what kind of protection may exist in different phases of the machine learning process.

The top layers of patents and trademarks, are to be applied separately and therefore can be described as layers that need to be actively built on top of the machine learning components.

Trade secrets, databases and copyright are protected, at least in Europe, automatically, and as seen in the graph, items belonging to these groups are used in several steps of the machine learning process.

And when we take a look at the items that receive no IP protection, the hybridity of AI becomes even more evident.

When we broaden the perspective to cover also items that are generated in the machine learning process, we can see that the emphasis of non-protected components increase.

The resulting graph maps the spectrum of IP protection to AI as a hybrid asset. It can be used when identifying for instance the legal and business needs or painpoints embedded in the machine learning process. In the following, we will use the graph to pinpoint certain aspects relating to AI.

2. Copyright and AI

3. AI as hybrid intellectual property

4. Principles for truly Open Source AI

5. Promoting openness to respect software freedom