"Connecting communities, empowering nations: Unifying the Regional and Global Economy ."

A small manufacturer may soon receive an AI tool at almost no licence cost and still find that using it creates an expensive dependence. The software may be open, while the computing power, maintenance, upgrades and practical knowledge remain elsewhere. For India, this is the uncomfortable question behind the next wave of international AI cooperation: how much capability will arrive with access?

China announced plans to lead an open-source AI ecosystem for BRICS, while India warned against the weaponisation of technology and critical minerals. These positions expose a central tension in technology cooperation: widening participation while protecting freedom of choice. The announcement is reported by “The Indian Express” (https://indianexpress.com/article/india/pm-modi-brics-speech-critical-minerals-weaponisation-supply-chain-startups-10876019/). The reporting reviewed does not establish a detailed technical architecture, governance model or binding access conditions. Any assessment of future dependence must therefore be treated as a risk to examine, rather than a conclusion about the proposed initiative.

An old development problem enters the digital age. Industrial history offers a useful warning. Buying a machine did not automatically create the ability to design it, repair it or produce its critical parts. Technology transfer delivered deeper benefits when local engineers and enterprises learned to adapt what they acquired. Where that learning remained weak, production expanded but dependence continued through spare parts, maintenance and specialised knowledge.

AI could reproduce this pattern with less visible machinery. A country may have thousands of businesses using advanced models while possessing limited ability to maintain the systems around them. Adoption figures could rise faster than domestic competence. Policymakers might celebrate the number of users while overlooking who earns the recurring income and who can change the terms.

There is a profound difference between an economy that uses intelligence and an economy that can shape how that intelligence works.

Open access needs an economic test. Open-source AI can provide real advantages. It can allow inspection, adaptation and competition among service providers. The Open Source Initiative defines openness through freedoms to use, study, modify and share, supported by access to the components needed to exercise those freedoms. A free interface alone does not establish those rights. “Open Source Initiative” (https://opensource.org/ai/open-source-ai-definition)

Even meaningful openness, however, does not remove every operating cost. Businesses still need suitable computing resources, organised data, integration with existing systems, security, training and someone who can fix failures. A model that looks inexpensive in a demonstration may become costly when it must work reliably every day.

Consider an illustrative calculation, not a market estimate. If an AI application saves a firm ₹15,000 a month but computing, support and checking its output cost ₹12,000, the benefit is only ₹3,000 before recovering the initial investment. If errors cause rejected orders, the saving may disappear altogether. The useful measure is the cost of a dependable business result.

Open models can reduce dependence when users can operate them independently and change service providers. Dependence may persist when those freedoms exist on paper but remain too expensive or technically difficult to exercise.

The strategic contest may move beneath the software. Giving away a model can still support a profitable business in hosting, hardware, implementation and associated services. That is not evidence of hostile intent. It is an economic reason to examine the complete arrangement.

The same scrutiny should apply to Chinese, American, European and Indian providers. Nationality alone cannot establish whether an enterprise retains control. An Indian intermediary could also make it difficult to move data or change suppliers.

Over the coming decade, influence may increasingly rest with those who establish the systems through which businesses purchase, produce, communicate and make decisions. Once employees are trained, production processes are connected and years of records accumulate, switching becomes harder. The cheapest entry point can become a costly exit.

The connection with critical minerals is therefore significant. Digital capability still rests on physical resources: chips, electricity, communications equipment and data centres. Software openness cannot by itself guarantee access to that underlying infrastructure.

For MSMEs, the missing institution is local implementation. A garment producer in Tirupur needs help identifying defects, estimating fabric requirements or responding to buyers. An engineering unit in Faridabad may need better maintenance planning. A food processor needs reliable records and production control. These enterprises require tools that understand their operations and people who remain available after the demonstration ends.

India should therefore treat AI adoption as an extension of industrial development. Cluster associations, technical institutions and business service providers could jointly organise implementation teams. Shared facilities could test applications, compare providers, train workers and negotiate support arrangements. Public assistance should help enterprises become capable users, with enough knowledge to question the advice they receive.

Shared services must also preserve commercial confidentiality. Firms should be able to obtain technical support without surrendering customer lists, product designs or pricing information to a common pool. Transparent data governance means clear answers about what is collected, where it goes, who can use it and how it can be removed.

Success should be assessed through lower rejection rates, shorter delivery times, reduced downtime and sustained gains after subsidies end. Counting workshops or registrations reveals little about whether production has improved.

Cooperation should expand the freedom to choose. India can engage constructively with a BRICS AI initiative while seeking published licences, practical options for domestic hosting, independent evaluation, shared governance and affordable movement between providers. These conditions would make cooperation more credible for every participating country.

Building domestic capability does not require producing every component at home. It requires understanding the dependencies, developing alternatives where they matter and retaining people who can adapt the technology.

The future danger is that India becomes an enormous market for AI applications while remaining a minor participant in their design, maintenance and economic returns. The future opportunity is to build a broad base of local firms that turn accessible models into useful industrial solutions.

A meaningful AI partnership should leave a small enterprise more capable each year. If its use of technology grows while its freedom to question, repair or change that technology shrinks, access has advanced further than development.

MSMEs #ArtificialIntelligence #BRICS #TechnologyPolicy