The Next Billion AI Users Won't Choose Between ChatGPT and Gemini.

Most coverage of Xi Jinping's BRICS AI pledge focuses on governance details like leadership and declarations, and its link to the World AI Cooperation Organization (WAICO) launched in July. However, the key issue is distribution. China isn't aiming to out-legislate the EU AI Act but wants to introduce an AI model to the next billion people unserved by current options, before OpenAI or Google do.

At the BRICS leaders' summit in New Delhi on 13 September, Xi proposed that China lead a BRICS "open-source zone for artificial intelligence": joint development of large language models, specialised AI research and training programmes, and an "open ecosystem" for the group's members, part of a wider "Greater BRICS Initiative" that also covers a special economic zone partnership and a shared digital cloud platform. BRICS, plus the partner countries that have joined since 2024, now accounts for roughly half the world's population, close to 3.9 billion people. A large share of that population has never paid for ChatGPT, never signed up for Gemini, and in some of those markets could not easily do either even if they wanted to: pricing in dollars, data localisation law, or outright restrictions on US cloud services all get in the way. Xi's pledge targets that gap directly.

BRICSS now accounts for roughly half the world's population, close to 3.9 billion people.

It is not a hypothetical gap. Chinese open-weight models have already done this once, fast. Chinese labs' share of global open-source model usage went from roughly 1 percent in late 2024 to close to 30 percent within about a year, on independent analysis of usage across large token volumes. Alibaba's Qwen family passed a billion downloads on Hugging Face by March 2026, faster than any open model family in history, and Chinese open models collectively passed 10 billion cumulative downloads by April, with four in five US AI startups now building on Chinese open-source models somewhere in their stack. None of that required a BRICS declaration. It happened because the models are free, strong, and easy to self-host, which matters enormously to a developer or a business in a market where a metered API in dollars is a real barrier. A state-backed "open-source zone" is that same dynamic with training programmes, cloud infrastructure and government backing layered on top of it.

four in five US AI startups now building on Chinese open-source models

For a Western AI company, that is a market-share story: users and developers who were always going to be hard to reach with a dollar-priced, US-hosted product now have a faster, cheaper, actively promoted alternative. For a multinational GRC team, it is a different kind of problem, and a more immediate one. Global organisations have not run IT from a single hub for years. Data residency law already forced that: GDPR in Europe, China's own data localisation rules, India's DPDP Act, and a growing list of sovereign cloud requirements mean a bank or an insurer with operations in a dozen countries is already running infrastructure in a dozen places, not one. India's own answer to this pressure, Sarvam, selected in 2025 to build a sovereign model under the IndiaAI Mission, markets itself as "sovereign by design," DPDP compliant, with air-gapped deployment available, exactly because global companies operating there need AI that satisfies rules a US-hosted model was never built to satisfy in the first place. A state-backed Chinese open-source ecosystem, actively distributed across BRICS markets, adds another branch to that same structure. It does not replace the AI stack a multinational already runs. It adds to it.

This is the practical outcome to consider. Even before Xi spoke in New Delhi, the list of AI vendors for a global organization was already diversifying due to regulatory pressures. A distribution campaign targeting billions who haven't yet chosen a primary AI vendor will accelerate that diversification, not hinder it, because it offers a credible, well-funded, actively promoted alternative to the markets driving fragmentation. A governance team expecting AI adoption to eventually consolidate around two or three well-known vendors should prepare for the opposite scenario. Governing this AI complexity will backfire on any CIO and CFO. 

Anove's insAIght platform is built for that reality-of-speed rather than against it: one register that tracks every AI system a global organisation uses, regardless of which country it was built in or which cloud it runs on, with the provenance, the controls and the evidence attached to each one and the potential risk it poses. Teams that want a quick read on what an AI tool already in use discloses about its own origins can run it through ExplAIn, which is free.

Learn more

• insAIght, Anove's AI governance and risk platform, for governing an AI vendor list that spans multiple ecosystems and jurisdictions from one register.
• Two Templates, One Rulebook: What WAICO Means for AI Governance Outside Europe, on the organisation this initiative builds on.
• ExplAIn, our free tool for checking what an AI system already in use discloses about itself.

If your organisation's AI footprint is already spreading across vendors and jurisdictions faster than your inventory can track it, book a demo and we will show you how insAIght keeps up.