Arguments about what is and isn’t “open source” are often resolved by deferring to the Open Source Initiative (OSI): If a piece of software is available under a license rubber stamped as “open source” by the OSI’s formal “definition,” then that software is open source.

But waters muddy when you get into the nuts and bolts of legal definitions versus the “spirit” of what open source really means. Indeed, there is significant nuance in the open source versus proprietary software debate: Has an “open source company” hamstrung its project by sliding core features behind a commercial paywall? How much transparency is there around the project’s development? And how much direct input does the “community” really have in a given project?

To many, open source is not just about the legal ability to use and modify code; the culture, transparency, and governance around it is paramount.

Everyone knows about the Google-flavored version of Android that ships on smartphones and tablets, replete with an array of apps and services. The underlying Android Open Source Project (AOSP), released under a permissive Apache 2.0-license, is available for anyone to access, “fork,” and modify for their own hardware projects.

Android, by just about any definition, is about as open source as it gets. And Google has used this fact in its defense against anti-competition criticism, noting that Amazon has reappropriated Android for its own lineup of Fire-branded devices. But all this ignores separate “anti-fragmentation agreements” Google signed with hardware makers that restrict them from using forked versions of Android. And unlike something like Kubernetes that sits under an independent foundation with a diverse range of corporate and community contributors, Android sits under the direct control of Google without a great deal of transparency over roadmap or community input.

“Android, in a license sense, is perhaps the most well-documented, perfectly open ‘thing’ that there is,” Luis Villa, co-founder and general counsel at Tidelift, said in a panel discussion at the State of Open Con25 in London this week. “All the licenses are exactly as you want them — but good luck getting a patch into that, and good luck figuring out when the next release even is.”

This gets to the crux of the debate: Open source can be something of an illusion. A lack of real independence can mean a lack of agency for those who would like to properly get involved in a project. It can also raise questions about a project’s long-term viability, evidenced by the countless open source companies that have switched licenses to protect their commercial interests.

“If you think about the practical accessibility of open source, it goes beyond the license, right?” Peter Zaitsev, founder of open source database services company Percona, said in the panel discussion. “Governance is very important, because if it’s a single corporation, they can change a license like ‘that.’”

These sentiments were echoed in a separate talk by Dotan Horovits, open source evangelist at the Cloud Native Computing Foundation (CNCF), where he mused about open source “turning to the dark side.” He noted that in most cases, issues arise when a single-vendor project decides to make changes based on its own business needs among other pressures. “Which begs the question, is vendor-owned open source an oxymoron?” Horovits said. “I’ve been asking this question for a good few years, and in 2025 this question is more relevant than ever.”

These debates won’t be going anywhere anytime soon, as open source has emerged as a major focal point in the AI realm.

China’s DeepSeek arrived with a bang off the back of open source hype, and while the models’ MIT licenses are very much recognized as open source, there remains black holes around training data among other components. Which is why researchers at Hugging Face are trying to create an even “more open” version of DeepSeek’s reasoning model.

Meta, meanwhile, has long tooted its open source horn with regards to its Llama-branded large language models (LLMs), even though Llama isn’t open source by most estimations — the models, while perhaps more “open” than others, have commercial restrictions.

“I have my quibbles and concerns about the open source AI definition, but it’s really clear that what Llama is doing isn’t open source,” Villa said.

Emily Omier, a consultant for open source businesses and host of the Business of Open Source podcast, added that such attempts to “corrupt” the meaning behind “open source” is testament to its inherent power.

“It goes to show how strong the brand of open source is — the fact that people are trying to corrupt it, means that people care,” Omier said during the panel discussion.

Much of this may be for regulatory reasons, however. The EU AI Act has a special carve-out for “free and open source” AI systems (aside from those deemed to pose an “unacceptable risk”). And Villa says this goes some way toward explaining why a company might want to rewrite the rulebook on what “open source” actually means.

“There are plenty of actors right now who, because of the brand equity [of open source] and the regulatory implications, want to change the definition, and that’s terrible,” Villa said.

While there are clear arguments for applying additional criteria that incorporates the “spirit” of what open source is intended to be all about, having clear parameters — as defined by a license — keeps things simple and less subject to nuanced subjectivity.

How much community engagement would be necessary for something to be truly “open source”? On a practical and legal level, keeping the definition limited to the license makes sense.

Stefano Maffulli, executive director at the OSI, said that while some organizations and foundations do lean into ideas around “open design, community, and development,” these are all fundamentally philosophical concepts.

“The point of having definitions is to have criteria that can be scored, and focusing on licensing is how that is accomplished,” Maffulli said in a statement issued to TechCrunch. “The global community and industry have come to rely on the Open Source Definition and now the Open Source AI Definition as objective measures that they can rely on.”