The open-source AI definition

What is open-source AI

An open-source AI is an artificial intelligence system made available under terms and in a way that grant the four freedoms derived from the Free Software Definition:

  • to use the system for any purpose and without having to ask for permission.
  • to study how the system and its components work.
  • to modify the system for any purpose, including to change its output.
  • to share the system for others to use with or without modifications, for any purpose.

These freedoms apply both to a fully functioning system and to each of its components.
A precondition to exercising these freedoms is to have access to the preferred form to make modifications to the system.

Preferred form to make modifications to AI systems

The preferred form for making modifications to an AI system is the one used by the developers of the system to create it and must include the elements below:

  • Source Data: A copy or pointer to the data used to create the system, so that they can be freely findable, accessible, interoperable, and reusable.
  • Source Code: A copy of the source code of each software used to retrieve the Source Data, to create, and to run the system. 
  • Processing Information: Detailed information about how the Source Data were collected and used to create the system, so that a skilled person can recreate a copy of the system from the Source Data and Source Code.
  • Parameters: The model parameters, such as weights or other configuration settings.

Each of these elements shall be available under terms and in a way that fully grant each and all of the four freedoms, e.g. dedications to the public domain, licenses that comply with the Open Source Definition and so on.

The licensing or other terms applied to these elements and to any combination thereof may contain conditions that require any modified version to be released under the same terms as the original.

Open-source models and open-source weights

The preferred form to make modifications to AI systems also applies to their individual components: “open-source models” and “open-source weights” must include the source data, the processing information, and the source code used to derive those parameters.


Version 1.0