Free and Open-Source Licensing and Regulation of AI Technologies (Part 3/3)

Free and Open-Source Licensing and Regulation of AI Technologies (Part 3/3)

In Part 2 of This Series…

…we discussed how free and open-source licenses (FOSL) intersect with current legal definitions and regulatory frameworks, particularly in the context of the EU AI Act. We also explored how FOSL can exempt certain AI systems and general-purpose AI models from regulatory obligations and examined the legal uncertainty surrounding intellectual property rights in AI components like AI models.

In this part, we now focus on available FOSL options for AI technologies and highlight key legal, ethical, and strategic considerations for developers and users when selecting and applying such licenses. Let’s start!

Introduction

Let’s say you are developing AI technologies. Perhaps you wish to simply use such technologies in your company. Or maybe you want to modify already existing AI components, adapting them to your needs and use cases. Why should you care about free and open-source licensing?

In this blog series, we address some key questions in an area that so far has not received much attention in the overall debate around AI regulation and adoption in Europe and elsewhere. We conclude that you should care about free and open-source licensing if, for example, you wish to share your technology but preserve your IP rights, while, at the same time, enforce some ethical restrictions. Or, if you want to integrate a component subject to an open license into your system. 

Why should you care about open licensing as an end-user?

You should care about open licensing also if you are an end-user of AI tools released under an open license. Public (open) licenses do not require individual contact with the rights holder. They offer an interesting avenue for both developers and users and bear important legal, contractual, and strategic implications for the parties.       

What is free and open-source licensing anyway, and what does it have to do with AI technologies? What are the implications of such licensing for my IP as a developer or my options as an end-user? What licensing platforms are available out there, and which one might be suitable for me? What are the advantages of subjecting my AI tool to an open license, e.g., in terms of compliance with the AI Act?

Our blog series will attempt to bring some clarity to this topic and help you understand whether free and open-source licensing is relevant for you, and why.

What Options of FOSL Are Available for AI Technologies? 

Assuming you are a developer, and you consider releasing your AI system or model under an open license: What options do you have?  

A partial list of conceptualization efforts and licensing platforms specifically designed to cover AI technologies includes the Open Source Initiative (OSI). This initiative offers its definition of Open Source AI with the understanding that it can foster safety & transparency, competition, polyculture, and application diversity. The OSI explains the distinction between this definition and the common open-source concept for software programs in the following way:

“The Open Source Definition (OSD) refers to software programs. AI and specifically machine learning systems are not simply software programs; they blend boundaries with data, configuration options, documentation and new artifacts, like weights and biases. The Open Source AI Definition describes the preferred form to modify an AI system, providing clarity on interpreting the principles of the OSD in the domain of AI.”

The Open Source AI Definition (AIOSD) echoes the longstanding open-source principle by allowing everyone to use, study, modify, and share (“freedom”). The OSI further distinguishes “open weights” licensing, a more limited permission which can be subject to various OSI licenses, from Open Source AI, that, by definition, mandates the sharing of the training code, training data composition, and possibly also the training dataset.   

Also on our short list is Creative Commons, which seems to have acknowledged that its licenses could be used particularly in connection with copyrighted works used to train AI models. (For a deep dive on this topic, you might want to check out this article). It might be less helpful when it comes to covering entire AI systems and/or their individual components. There is, of course, a fundamental difference between the two types of inquiries: Copyright infringement during model training vs. copyright infringement that happens while using the (already trained) model. At the same time, a CC license can still apply to a copyrighted software (for example, the inference system code that executes instructions encoded by the model weights) or databases included or designed as a component of an AI system.

Introducing Responsible AI Licenses (RAIL)

Another, more recent initiative that specifically aims to empower AI developers in connection with their IP in technology is the Responsible AI Licenses (RAIL). It aspires helping developers to solve the following moral dilemma: On the one hand, they wish to share algorithms, code, or data, but on the other hand, they worry that someone will use this knowledge for purposes they consider irresponsible or harmful.

RAIL explains on its website:

“These licenses include behavioral-use clauses which grant permissions for specific use-cases and/or restrict certain use-cases. In case a license permits derivative works, RAIL Licenses also require that the use of any downstream derivatives (including use, modification, redistribution, repackaging) of the licensed artificial must abide by the behavioral-use restrictions.”

RAIL offers a modular structure that allows developers to compile a license according to their preferences, but it also offers ready-to-use templates. Interestingly, there is a designated license for AI models (AIPubs OpenRAIL-M), a separate license for source code (AIPubs OpenRAIL-S), as well as end-user license agreements and licenses covering other components such as data and applications. The licenses provide important definitions to key concepts that can certainly contribute to the development in this area. For instance, in the AI Pubs Open RAIL for models, we can find designated definitions to terms like “Model”, “Complimentary Material”, and “Derivatives of the Model”.

Also here, the power of these legal instruments is significantly derived from their direct reliance on IP as the licensing baseline for AI technologies. For example, the RAIL license for models mentioned above states under Section II that:

“Both copyright and patent grants may apply to the Model and Derivatives of the Model. The Model and Derivatives of the Model are subject to additional terms as described in Section III, which shall govern the use of the Model and Derivatives of the Model even in the event Section II is held unenforceable.”

Accordingly, this instrument is designed to provide a public copyright and patent license alongside a list of use/behavioral restrictions specified under its “Attachment A” concerning harm, discrimination, and transparency. The intention is that the license will operate to some extent similarly to other open-source licenses, yet with a strong(er) focus not only on freedom but also on ethical restrictions such as “do-no-harm”. For more details, you might want to read this blog article.

What Does FOSL Mean for Me as a Developer?

As an AI developer, you now better understand FOSL and can turn to key questions: Which licensing model fits your goals? What trade-offs are you willing to make? Image created with MidJourney.

As a developer of AI technologies, you now understand better what FOSL means and how it could promote your objective. You may now turn to focus on the questions that really matter for you and make an informed decision about your licensing strategy.

  • Exclusive rights: What IP covers my technology (if any)?
  • Licensing strategy: Do I want to pursue a proprietary or rather an open licensing strategy?
  • Subjective-ethical preferences: Do you endeavor to promote, through a conscious licensing choice, the freedom to use and exchange information, to collaborate, study, share, and build on existing knowledge? Do you want to impose behavioral restrictions?
  • Objective-legal preferences: Are you willing to grant public access and use rights regarding your IP, knowledge, and data, and if yes, under which conditions?
  • Share-alike: Do you insist that all users and modifiers of your technology on the value and distribution chain respect your terms?
  • Monetization: Do you want to directly or indirectly (e.g., via services, support, or complementary products) monetize your technology? Do you want to restrict commercial use in the license?
  • Regulative advantages: In terms of compliance with regulations covering AI systems and models, do you stand to gain any advantage by choosing a certain FOSL?
  • Simplification and cost-effectiveness: Can you simply use available “off-the-shelf” FOSL, or freely available modular instruments that you configure and adopt without any legal costs for drafting the license?
  • Support: Do you need specialized assistance in answering all these questions, or can you solve them in-house?

Regarding the last question, it is likely that specialized knowledge would be required in the following areas:

  • Identifying IP rights and exclusivity domains;     
  • Identifying the range of available public licenses that can best reflect your subjective preferences and objectives;
  • Identifying the need to adapt the legal text of the license to your individual needs;
  • Understanding the legal and business consequences of subjecting your technology to a certain license;
  • Identifying risks and opportunities of a given licensing strategy;
  • Understanding the consequences of breaching the terms of the license and enforcement.

What Does FOSL Mean for Me as a Downstream User?

If you are not the original developer who holds the rights in the AI technology but you want to use it for your purposes, whether for-profit or not, whether as an end-user or as a modifier that offers enhanced products to its customers, the principal remains the same: You must understand from the outset the licensing terms that apply to the technology, what is covered by the license and what not, and what do you need to do if you want to use the technology at minimum or no legal risk.

This review process, which typically is part of the compliance/governance scheme of the organization, applies to the various stages of strategic planning, product ideation, development, deployment, integration in other systems/products, modification, and redistribution. It is helpful to be able to identify technologies and tools released under FOSL, in many cases free of charge, which you can utilize and deploy. It is further essential to understand the limits of the FOSL to avoid liability risk and financial loss down the road.

Conclusion

Free and open-source licensing (FOSL) represents a crucial, yet rather complex, legal and strategic dimension for the development, deployment, and use of AI technologies. As explored in our blog series, FOSL offers developers an avenue not just to share their innovations, but also to retain some control over intellectual property and apply ethical usage restrictions, when desired. For AI users and organizations, FOSL arrangements hold the promise of enabling broader access to advanced technologies, reducing entry barriers and transaction costs, and promoting collaborative innovation – provided the licensing terms are thoroughly understood and respected.

What does the EU AI Act really mean for open-source AI?

The recent EU AI Act and associated guidelines concerning general-purpose AI models provide important clarification on the regulatory treatment of open-source AI, carving out some limited exemptions for FOSL systems – except in higher-risk or commercialized contexts. This regulatory stance acknowledges the value of FOSL in promoting research, innovation, and transparency. However, the regulatory landscape is still evolving, with ongoing EU consultations that are likely to shape future guidelines and the Code of Conduct, possibly around the interplay of open-source licenses with ethical restrictions and monetization.

What developers and users should prioritize

For AI developers, navigating FOSL requires a careful assessment of their technology’s IP status, their licensing objectives (including whether to pursue openness, commercialization, or ethical controls), and the practical implications of different available licenses – be it GPL, Creative Commons, or newer solutions like RAIL that are specifically designed to cover AI technologies. Users, in turn, must prioritize compliance and risk management, since a misunderstanding of licensing terms can carry significant legal and financial consequences.

Outlook: Is open-source AI heading toward mainstream adoption?

Looking ahead, a convergence between AI technology, regulation, and open-source might pick up on relevance and importance in the future. These licenses for AI are yet to be tested in courts, and there are hitherto only few attempts to enforce them, possibly due to the complexity of the legal questions. However, we can anticipate increasing specialization of FOSL instruments for AI, further regulatory clarity, and a growing culture of openness balanced with accountability. Both developers and users must stay informed and agile as the landscape evolves, ensuring their innovation and adoption strategies align with legal and ethical best practices and with their own risk appetite.


Author info

Back to top
applydata garage barcamp 2025