Open-source software has long been foundational of the digital economy. By allowing unrestricted access, modification and redistribution of source code, open-source licensing is an enabler for collaborative innovation across industries, from software infrastructure to artificial intelligence.
However, a growing number of technology companies and research institutions are considering a different approach: source-available licenses. These licenses make the source code publicly accessible but impose certain restrictions on its use, typically limiting commercial exploitation or specific deployment scenarios.
Distinguishing Source-available licenses
At first glance, source-available software may resemble open source. The code is visible, accessible and often distributed through the same channels used for open-source projects, such as public repositories. The legal distinction, however, is significant. Under the Open-Source Definition, a license must not restrict fields of use or commercial exploitation. Source-available licenses depart from this principle by allowing access to the code while maintaining certain proprietary controls.
This distinction becomes clear when examining typical licensing clauses. Many source-available licenses explicitly restrict commercial use unless a separate commercial agreement is obtained. For example, clauses similar to the following are frequently found in source-available frameworks:
“Use of the Licensed Work in production requires a commercial license from the Licensor.”
Such provisions prevent the license from qualifying as open source, since they impose a restriction on commercial activity. The software therefore remains accessible for study, experimentation or limited use, but its commercial deployment remains under the control of the licensor.
Use and take up
This model has gained traction particularly among venture-backed technology companies, where licensing is increasingly viewed as a strategic component of intellectual property protection. Traditional open-source licensing can enable rapid ecosystem adoption, but it may also allow large competitors, including hyperscale cloud providers, to commercialize the technology without contributing proportionally to its development. Source-available licenses aim to address this tension by maintaining transparency and community engagement while protecting the core commercial value of the technology.
Several companies have adopted this approach in recent years, including Elastic and MongoDB, which moved away from open-source licensing to protect their technology from commercialization by large cloud providers. Licensing models such as the Business Source License (BSL)[1] or the Functional Source License (FSL)[2] publish the source code while restricting certain commercial uses for a defined period. Others rely on additive mechanisms such as the Commons Clause[3], a restriction that can be appended to an existing open-source license to prohibit the commercial sale of the software as a service, effectively converting it into a source-available model without replacing the underlying license. Similarly, initiatives such as the Source Available License proposed by Open Core Ventures[4] seek to provide a clearer and simpler alternative to fully closed-source models while preserving commercial defensibility.
In this sense, source-available licensing represents an intermediate model between open source and proprietary software. The objective is often not to prevent collaboration, but to structure it in a way that preserves sustainable business models. From a venture capital perspective, this approach can help maintain technological differentiation while still benefiting from community visibility and developer engagement.
Source-available licensing may be particularly well suited to certain recurring scenarios. This is often the case where broad visibility of the code is commercially useful, but unrestricted third-party exploitation would undermine the licensor’s ability to capture value. Typical examples include infrastructure software developed by venture-backed companies that wish to encourage developer adoption while preventing direct commercialization by large cloud providers (Mongo and Elastic are examples of this); AI systems or model-adjacent tooling where transparency supports trust, testing or ecosystem uptake, but unrestricted production use would erode the commercial advantage of the original developer (for example Mistral AI, which releases model weights publicly under research licenses permitting testing and non-commercial use, while requiring a separate commercial license for any self-deployment in production)[5]; and research-based technologies that benefit from early scientific access while preserving downstream licensing, spin-off creation or industrial partnerships (e.g. Academic Non Commercial licenses). In such contexts, source-available terms may offer a workable balance between openness, market positioning and control over monetizable uses.
Trade-offs
That said, this model is not without trade-offs. Where licensing terms depart from standardized and widely recognized open-source frameworks, potential adopters, customers, investors or commercial partners will often need to carry out a more detailed legal assessment before accepting them. In practice, this may create friction in procurement and technology adoption processes, delay negotiations and increase transaction costs. That friction may be compounded from a regulatory perspective: where the relevant framework does not qualify as free and open-source software, the licensor may also fall outside certain exceptions or lighter regulatory treatments available under EU instruments such as the AI Act and the Cyber Resilience Act. In that sense, greater commercial control may come at a double cost: reduced transactional simplicity and reduced access to regulatory carve-outs designed for genuinely open ecosystems. For that reason, where a source-available approach is preferred, relying on licensing models with a higher degree of market familiarity — such as the BSL or, increasingly, the FSL mentioned above — may help mitigate interpretative uncertainty, even if it does not remove the underlying regulatory distinction.
Particularly harmful have been technology licensing models claiming to be open source, but actually with some “source-available” style restrictions. A growing concern in the technology and AI industry is the practice of “open washing”: the use of open-source language to describe licensing models that, in practice, retain significant restrictions on use. As Dries Buytaert, founder of Drupal, argued in December 2025, the term “open source” has a specific, shared meaning shaped by decades of community work — and repurposing it for marketing, however well-intentioned, creates real confusion in the market and ultimately undermines the credibility of genuine open-source ecosystems.[6] The most prominent example is Meta’s series of Llama large language models.
Despite Meta marketing Llama as open source, the Open-Source Initiative (OSI) formally stated that the Llama license does not meet the Open-Source Definition, since it discriminates against certain persons and fields of endeavor, conditions expressly prohibited by the OSI’s standards.[7] Specifically, Llama 2’s license required that any company with over 700 million monthly active users obtain a separate license from Meta, and prohibited the use of Llama outputs to improve other large language models. With Llama 4, the OSI reiterated that the model remains not open source, noting that its license excludes users domiciled or operating in the European Union. The pattern illustrates a broader risk: where source-available or otherwise restricted models are publicly described as open source, the resulting ambiguity can mislead developers, customers and investors, and ultimately undermines the credibility of genuine open-source ecosystems. Organizations adopting or building AI models should therefore look beyond marketing claims and assess the actual licensing terms before drawing conclusions about the degree of openness involved.
Available source in academia and for AI
Beyond the commercial context, source-available models are also gaining relevance in academic and research environments for distinct institutional reasons. Unlike venture-backed companies, universities and research institutions are often subject to public funding obligations, open science mandates and technology transfer frameworks that create competing pressures: an expectation of broad knowledge dissemination on one hand, and a duty to preserve the commercial potential of research outputs — through licensing, spin-off creation or industrial partnerships — on the other.
Licensing frameworks such as the Mass General Brigham licenses[8] illustrate this approach. By allowing access to research outputs under controlled conditions, these license models enable scientific collaboration while safeguarding the possibility of future commercial exploitation. In practice, they can support technology transfer processes in which early-stage research remains accessible to the scientific community without compromising potential industrial partnerships or spin-off creation. BrainIAC project[9] is an example of this.
These developments reflect a broader evolution in the way organizations approach intellectual property in digital and research-driven environments. Rather than choosing between fully open and fully proprietary models, many organizations are exploring hybrid strategies that combine transparency, collaboration and commercial protection.
Artificial intelligence is making this discussion even more relevant. As companies increasingly develop and deploy AI systems built on a combination of proprietary models, open-source components, training pipelines and curated datasets, licensing strategy becomes a central element of legal and commercial design rather than a secondary compliance exercise. In this context, source-available models may offer a particularly valuable framework: they can support transparency, foster developer adoption and research collaboration, while preserving control over high-value commercial uses, model deployment and downstream exploitation. For organizations operating in the AI space, the key issue is therefore not only whether technology should be accessible, but under which legal conditions that accessibility can coexist with defensibility, investment readiness and sustainable monetization.
Strategic considerations
For companies developing software platforms, AI systems or digital infrastructure, the key questions are no longer limited to whether code should be published or kept proprietary. Instead, organizations increasingly need to consider:
- whether community collaboration is essential for technological adoption
- how to prevent unintended commercial replication of core technology
- how licensing choices affect investment attractiveness and venture financing
- how research collaboration can coexist with sustainable commercialization
In this context, source-available licensing should not be understood simply as a legal technicality. It is increasingly becoming a strategic instrument for managing innovation, intellectual property and market positioning.
As the licensing landscape continues to evolve, with new models emerging alongside traditional open-source licenses, organizations releasing software, AI systems or research outputs should carefully assess the legal and commercial implications of their licensing choices.
If your organization is considering releasing technology under an open-source or source-available license, obtaining expert advice at an early stage can help ensure that the chosen licensing strategy aligns with both regulatory requirements and commercial objectives.
At Across Legal, we assist companies, startups and research institutions in designing licensing strategies that support innovation, collaboration and sustainable commercialization.
[1] MariaDB Corporation Ab, Business Source License 1.1 (BSL 1.1), https://mariadb.com/bsl11/. https://en.wikipedia.org/wiki/Business_Source_License.
[2] Sentry / Chad Whitacre, Introducing the Functional Source License, https://blog.sentry.io/introducing-the-functional-source-license-freedom-without-free-riding/; Functional Source License (FSL).
[3] https://commonsclause.com/
[4] https://github.com/OpenCoreVentures/open-core-ventures-source-available-license
[5] https://mistral.ai/news/mistral-large-2407
[6] https://dri.es/source-available-is-not-open-source-and-that-is-okay.
[7] https://opensource.org/blog/metas-llama-license-is-still-not-open-source
[8] https://github.com/MGB-Open-Science-Program-Office/open-science-license-library
[9] BrainIAC: A foundation model for generalized Brain MRI analysis – online at https://github.com/AIM-KannLab/BrainIAC
↳ Discover more content in the section Insight.
↳ Do you want to stay up today about the sector related news? Follow us on LinkedIn.
