The executive director of the Linux Foundation today suggested the consortium will soon be playing a bigger role in ensuring that data sources remain open in the age of artificial intelligence (AI).
Speaking at the Open Source Summit, Jim Zemlin told conference attendees that data is increasingly being moved behind a moat that limits access. That shift, however, will drive a counter reaction, he adds. The Linux Foundation sees an opportunity to play a larger role in making sure that open data remains accessible to AI agents and applications, says Zemlin.
“Bookmark this spot,” says Zemlin. “There is a lot of room here for the Linux Foundation.”
In general, AI models have already consumed much of the public data that is available. There may be room to establish any number of public-private partnerships to provide access to additional data, but more organizations are clearly looking to make the data they create available to AI models for some type of fee. While the terms of those fees may vary widely, the end result is the cost of training AI models is starting to increase.
At the same time, however, frontier AI models based on open source licensing models will start to have a much bigger impact in 2026, predicts Zemlin. In combination with the open sources of data there is a clear opportunity to reduce the total cost of AI, he adds.
Historically, the open source community has tended to lag behind whenever there is major new innovation. While open source has been more of a driver of innovation in recent years, AI models are being advanced today mainly by providers of proprietary platforms. As a result, the price of the token to provide inputs and generate outputs is set by a relatively small number of players. The Linux Foundation is now clearly signaling its intention to foster adoption of open source AI models trained using open data to lower the total cost of AI.
In fact, the Linux Foundation is reminding developers to be “promiscuous” when it comes to adopting AI models to prevent their organizations from becoming locked into a specific platform, says Zemlin.
Where the open source community can fulfill the promise remains to be seen. However, it’s already apparent that not every AI agent or application requires access to the most advanced AI models that, given their size and reasoning capabilities, are also the most expensive to use.
Regardless of the AI model used, there is little doubt that open source options are starting to proliferate. In addition to the controversial DeepSeek AI model, other options include Kimi from the Moonshot project, GLM from Zhipu AI, Qwen from Alibaba and Gemma-4 from Google.
It’s not clear at what pace these AI models are being advanced compared to offerings from OpenAI and Anthropic, but if history is any guide, the gap will only continue to narrow as more developers become familiar with the fundamentals of building and maintaining frontier AI models.
