The strategic imperative of open source AI

This year’s open-source pivot is the most consequential AI development since ChatGPT’s launch in November 2022. On 20 January, DeepSeek released its open-source reasoning system, an ‘AI Sputnik moment,’  in Marc Andreessen’s words. Then, on 23 July 2025, the Trump administration made open-source a US strategic priority in ‘Winning the Race: America’s AI Action Plan.’

This AI shift has been counterintuitive. Chinese companies historically favoured proprietary software, and Republicans were rarely open-source champions. Why did both powers shift course? Because open-source has moved from an ideological preference to a strategic priority.

From ideology to strategy

Open-source emerged from a 1960s Californian ethos favouring inclusion, transparency, and communal development, often associated with a leftist, ideological undertone. Today, its relevance is strategic. As Eric Schmidt, former leader of Google, said recently:

China is winning the AI race because it embraced the open-source approach. Countries around the world are more likely to accept Chinese open-source models than US closed and proprietary AI systems.

This logic poses a direct challenge: if Washington clings to closed systems while Beijing advances open ones, global AI standards could tilt toward China, mimicking what happened with internet protocols and Linux.

History’s pattern: Openness wins over proprietary solutions

Digital history offers a clue to how the current contest between open and proprietary AI platforms will develop. Namely, open source systems have repeatedly triumphed over proprietary ones.

The protocol war (TCP/IP vs. OSI)

From the 1970s to the early 1990s, a battle raged over the fundamental language of computer networking. The Open Systems Interconnection (OSI) model was a complex, top-down framework designed by European telecoms. Its rival, the Transmission Control Protocol/Internet Protocol (TCP/IP), was a pragmatic, open suite developed by a US research community. Though governments initially mandated OSI, TCP/IP won decisively because it was simpler, more useful, and freely available. It became the backbone of the internet.

The historical misnomer of openness is repeating itself. During the technology protocol wars, OSI (Open Systems Interconnection) failed to defeat TCP/IP largely because its standards were shaped by closed committees and implemented in proprietary systems.

Today, OpenAI presents a parallel case; despite its name, the company operates a highly proprietary AI platform, making its designation a significant misnomer. This contrast became especially clear recently when, following the launch of new US AI safety strategies, OpenAI’s shift away from its original open principles was highlighted.

The operating system war (Linux vs. Proprietary OS)

A similar dynamic unfolded in the 1990s as expensive, proprietary systems like commercial Unix and Microsoft Windows dominated the market. The challenger was Linux, a free, open-source operating system designed in 1991 by Finnish developer Linus Torvalds and developed by thousands of programmers and volunteers for the last 34 years.

Its collaborative model enabled blistering evolution, allowing it to dominate servers, cloud computing, and the Android ecosystem. Linux outcompeted rivals on cost, flexibility, and the relentless innovation of its open community.

These cases offer a clear lesson. Today’s powerful but proprietary AI models are the modern equivalent of OSI and commercial Unix. Open-source models like Meta’s Llama and DeepSeek are the TCP/IP and Linux of our time: accessible platforms that catalyse a broad ecosystem of innovation.

Table 1: A Timeline of Tectonic Shifts in Tech Paradigms

What are the strengths of open-source AI

Speed

Open models evolve at a pace that closed, corporate labs cannot match. They are continuously battle-tested and improved by a global community working in parallel. With over 1.2 million models on Hugging Face,  this decentralised approach allows thousands of contributors to simultaneously fix bugs, enhance capabilities, and unlock new applications. 

Cost

The open-source model fundamentally changes the economics of AI development. Costs are distributed across a vast community of contributors, researchers, and companies, drastically lowering the financial barrier to creating cutting-edge technology. This efficiency is amplified by a focus on smarter architecture over brute force. The AI race is not being won solely by those with the most GPUs.

Motivation

Open source ethos attracts younger developers who cherish impact, learning and community work. Surveys show Gen Z and millennials consistently value purpose and meaningful work, while open-source research communities report learning/networking as top motivators to contribute. That purpose-driven pull is a durable talent magnet for open ecosystems.

Integration

As large language models (LLMs) begin to hit scaling plateaus, the frontier of innovation is shifting from raw model size to integration in the economy and education. The true value of AI will be unlocked by embedding it into countless workflows across industries.

Open platforms are inherently better suited for this future. Their transparency and adaptability make them far easier to customise, audit for safety, and embed into diverse applications. This is evident in the explosive growth of open-source AI agents (e.g., frameworks like DiFy, N8N, LangChain) and specialised tools. This fosters a vibrant and diverse ecosystem that grows exponentially, creating a powerful network effect that closed, proprietary systems cannot achieve.

Size

Recent breakthroughs demonstrate that elegant design and algorithmic efficiency can produce smaller, more nimble models that rival or even surpass their larger, more expensive counterparts. For instance, the Mistral 7B model from Mistral AI delivers performance comparable to much larger models, emphasising that small can be smarter. This trend democratizes access to powerful AI, allowing organisations and individuals to run sophisticated models without the hardware requirements of massive proprietary platforms like GPT-4.

The ascendant open source paradigm

We are witnessing a major shift towards acceptance of open-source AI as not only ethically favourable but, even more importantly, a strategically superior solution for the future of AI.

Meta’s Chief AI Scientist, Yann LeCun, captured this shift clearly. Responding to those who see DeepSeek’s rise as ‘China surpassing the US,’ he argues they are misreading the trend. The real story is that ‘open-source models are surpassing proprietary ones.’

The direction is clear. Just as TCP/IP and Linux became the foundational layers of our digital world, open-source AI is emerging as the next dominant paradigm. Openness wins not because it is a moral ideal, but because it delivers superior interoperability, resilience, and innovation at scale. The AI open-source shift provides hope that societal values can be aligned with corporate interests.