Open source a ‘force multiplier’ for AI innovation

The world of artificial intelligence (AI) is increasingly being shaped by open source, which has been driving progress and innovation in the field, according to Jim Zemlin, executive director of the Linux Foundation.

During the opening keynote of KubeCon + CloudNativeCon China 2025 in Hong Kong this week, Zemlin, joined by Chris Aniszczyk, chief technology officer of the Cloud Native Computing Foundation (CNCF), noted the pivotal role that open-source models and infrastructure are playing in AI.

Zemlin pointed to the surprise release of a high-performing open-source large language model (LLM) by DeepSeek in January 2025 as a pivotal moment, noting that “it really showed that open source is a force multiplier, that the sharing of ideas accelerates innovation.”

What followed was the release of powerful open-source models every few weeks, prompting even leaders of proprietary AI companies like OpenAI’s Sam Altman to reconsider their stance on open source, he added.

Zemlin pointed to the financial and strategic value of open source, citing a study with Harvard Business School that rated the demand-side value of open source at about $9tn. “That’s why open source is so powerful, it’s freely available high-quality software that you use to build technology products and services,” he said, noting that this allows developers to focus on the value they provide to their customers, as around 70% of a modern product or service is already comprised of open-source components.

Zemlin identified agentic AI as the next frontier for open source innovation, describing AI agents as the technology that connects LLMs to real-world tasks and systems. An emerging stack of open-source tooling is developing around AI agents, with initiatives like model context protocol (MCP), a technology from Anthropic, and Agent2Agent (A2A) from Google, which are creating open standards for connecting AI agents to enterprise systems and to each other.

“This is going to be something that you’ll hear more and more about throughout this year, and you’ll see a whole new, emerging open source ecosystem around it,” said Zemlin, adding that this will lead to “better agents that are transparent, safe and effective”.

Building on the AI momentum, Aniszczyk announced a new initiative to bring the success of Kubernetes to AI, with the CNCF kicking off an effort to define what it means to run conformant AI workloads on Kubernetes. “What we’ve done with Kubernetes in conformance needs to also happen in the AI space, so we can have a standardised way to do inferencing and distributed models,” he said.

To address the licensing challenges of AI, the Linux Foundation has also introduced the Open Model, Data and Weights (OpenMDW) licence, a permissive licence specifically designed for AI models and their artefacts to provide clarity and consistency for enterprises.

“In the case of AI, you have data and weights, data cards and other components that have a different kind of licensing regime,” Zemlin explained during a media briefing on the sidelines of the conference. “And what we created was a way to essentially assess all of those components to understand, when you’re consuming an open source model, what you’re really getting.”

Amid the growing momentum around open source AI, Zemlin reminded developers that “open source really works when you build a community around it”, a point Aniszczyk underscored when he provided an update on the growth of the CNCF.

According to Aniszczyk, the CNCF now has about 287,000 contributors worldwide across more than 200 projects. China, in particular, has become a major force, consistently ranking as the second or third largest contributing country to CNCF projects. He highlighted several successful projects born in China, such as Dragonfly and Volcano, which are now used globally.

Despite global political tensions, Zemlin remains confident in the non-partisan nature of open source. “The good news is that open source software is exempt from export control rules, and therefore is something that we still want to continue to collaborate on,” he said.