Why open source matters for AI’s next phase
From the earliest days of computing, the ambitions of AI and the open source movement have been intertwined. The founder of the Free Software Foundation came from an AI lab, and the term "open source" was coined by the leader of an AI-focused think tank. Now that deep learning has pushed AI to the center of software development, that connection deserves closer scrutiny.
Open source influences AI development in three distinct ways. First, the dominant AI frameworks—PyTorch being the obvious example—are open source. So are essential tools for responsible AI, such as InterpretML for transparency and AIF360 for bias measurement.
Second, open source collaboration is shaping how AI systems get built. Trained models are increasingly released under open source or other public licenses, allowing developers to use, train, modify and redistribute them. This mirrors the library and package ecosystems that underpin modern software. The licensing choices vary widely: EleutherAI uses open source licenses for both tooling and models, while other projects adopt licenses that grant broad permission to use, modify and share models, but restrict uses their authors deem unethical.
Third, the ethos of free and open source software has inspired parallel movements in open access, open data and free culture. That "information commons"—Wikipedia, Common Crawl, the Pile and similar resources—has been foundational to AI progress. Without it, AI training would have been slower and confined to the largest organizations with proprietary data.
How AI is reshaping software development itself
AI is quickly changing the practice of software engineering. Code generation systems now act as pair programmers, helping developers write code faster. Since GitHub launched Copilot, similar systems have appeared from Amazon, Carnegie Mellon, DeepMind, Meta, OpenAI, Replit and Salesforce. Beyond generating new code—especially boilerplate and tests—these tools assist with documentation and code translation between programming languages.
The upside extends beyond convenience. AI-powered tools could make advanced techniques like formal methods more accessible, lower the barrier for new developers, and improve software quality while reducing cost. That is the same pattern open source has followed for decades: lower costs increase demand and overall opportunity for developers.
But AI also introduces new complications for software correctness and supply chain security. Models are opaque, and as they become dependencies, they raise fresh questions about provenance. The lessons learned from securing traditional software ecosystems—global, open collaboration among developers, security researchers and society—will need to be applied here as well.
Stewardship questions for the open source community
Developers may be at an inflection point comparable to the early web. Then, wider access to libraries and code accelerated open source but also forced FOSS communities to reconsider user autonomy and transparency as software moved from distribution to service. AI may prompt a similar shift in the economics of software, expanding advocacy from copyright licenses to public regulation.
At the center of this conversation is the Open Source Initiative, steward of the Open Source Definition. OSI is uniquely positioned to lead the discussion, and its Deep Dive: AI event is designed to tackle questions that lack settled answers.
One fundamental question is what "open source" even means for an AI system. Is a pretrained model its own preferred form for modification? What minimal set of precursors must be licensed openly for the model itself to qualify? How should training sets containing personal or sensitive data be handled? These questions matter beyond community boundaries—drafts of the EU AI Act reference open source AI systems without a clear definition of the term.
A second set of questions concerns governance. At what layers—technical, community norms, standards, legal, public policy—should open source AI projects be governed? Can different approaches be complementary, or do they inherently limit collaboration and interoperability?
The macro questions are equally important. How can open collaboration reduce the risks of military competition or technological surprise, while increasing the odds that AI helps address challenges like climate change? The open source ecosystem has already grown from individual labs to global communities to critical infrastructure backed by corporate investment, with governments now entering the picture. AI dramatically expands what software will mean to society, making it a good moment for open source communities to ask the hard questions that have long been simmering beneath the surface.
That is the conversation OSI is convening—one that goes well beyond the practical origins of open source, yet stays true to its founding aspirations. The hope is that the community, including its corporate sponsors, will participate fully.
