How privacy and sovereignty could push the adoption of Decen…

A handful of tech giants, with their massive datasets, billion-dollar GPU clusters, and global reach, seemed to control the Artificial Intelligence space, as they built the most powerful AI systems and dictated the rules of engagement. But only until decentralised AI came into picture.   

A quiet revolution is underway, driven not from Silicon Valley boardrooms, but from decentralised networks spread across hospital servers, local bank branches, and even idle machines in small towns.

Once an academic curiosity, decentralised AI is fast becoming a strategic lever for startups across sectors. By distributing training across devices, nodes, and communities, it promises to democratise AI development, protect sensitive data, and help nations like India assert technological sovereignty.

Why Decentralisation Matters

“Decentralised AI lets sensitive sectors like healthcare or finance keep data where it belongs, in hospitals, banks, or government servers, rather than sending everything to a central cloud. That makes it easier to comply with privacy laws and build public trust,” explains Shashank Sripada, cofounder & COO of Gaia, a company that decentralises AI Inferencing.

Sripada notes that decentralisation allows AI models to be trained on local data under local rules, ensuring sovereignty. “In practice, this is how nations can claim technological sovereignty in AI: you benefit from global innovation but don’t hand the keys to a few foreign companies”.

This message resonates strongly with policymakers and enterprises wary of centralised monopolies. Hitesh Ganjoo, CEO of Iksha Labs, highlights how India’s hospitals and banks are already experimenting with this model. “We implemented [decentralised AI] in a leading private bank, where customer risk models were trained across regional servers without ever pooling raw financial data, and in a multi-city hospital network, where diagnostic models were trained inside hospital firewalls via federated learning. Both reduced regulatory risk while increasing adoption.”

Levelling the Playing Field for Startups

Big Tech has long dominated AI due to its control over data and GPU clusters. Decentralisation could change that.

Ganjoo echoes this, pointing to live pilots: “Regional banks are contributing to fraud detection models without exposing raw transaction data. Edge AI brings lightweight, domain-tuned models directly into hospital IT environments. Blockchain-based AI adds auditability, which could help startups establish credibility in BFSI and healthcare, where trust is everything.”

For legal experts, this levelling effect is just as critical for competition policy. Shivanghi Sukumar, partner at Axiom5 Law Chambers, argues: “Centralised AI models require capital investments in expensive cloud infrastructure, a significant barrier for Indian startups. Decentralised AI counters this by allowing companies to access a distributed network of GPUs and other computational resources. This lowers the financial entry barrier, aligning with the goals of the government’s IndiaAI Mission.”

EdgeUp by Zaryah Angels is an AI-driven ed-tech platform for UPSC coaching, utilising a proprietary small language model trained on local datasets. The founders are reportedly exploring decentralised AI computing to cut infrastructure costs. 

Indrajaal by Grene Robotics is an autonomous drone defence system that employs edge AI and distributed command/control nodes, operating in a decentralised manner.

The Emerging Business Models

If compute and data are the fuel, marketplaces are the new engines. Shibu Paul, VP of International Sales at Array Networks, notes that decentralised AI marketplaces will transform collaboration.

“Instead of depending only on hyperscalers, smaller businesses and even individuals with idle capacity will be able to contribute. This mirrors the early days of cloud adoption, but on a global, distributed scale,” Paul said.

The model extends to data. Decentralisation enables secure pooling of specialised datasets, with contributors rewarded via credits or tokens. “Communities of small and medium businesses could train models tailored to their sectors, such as healthcare, agriculture, or transportation. Instead of surrendering control of proprietary data to central authorities, they would retain ownership while collectively benefiting from the trained models”.

Over time, Paul believes these marketplaces will bundle compute, data, and models into industry-specific solutions, creating a transparent and equitable digital economy.

Challenges Ahead

Yet decentralisation is not without its hurdles.

The hard part is stitching all this together at scale, said Sripada, adding, “training across hundreds of nodes is messy; networks fail, updates lag, and models can be biased if each data source is too narrow. Security is another big one: you need to make sure no one is poisoning the model or leaking information through updates.”

Expanding on the pain points, Ganjoo said that most frontier models today are not designed for decentralised deployment. “Optimising them to run ‘small and local’ is still a frontier problem. Enterprises also lack awareness of how decentralised frameworks solve security and compliance problems. Education and ecosystem evangelism are as important as the technology itself.”

There’s also the tricky economics of adoption. While decentralisation reduces dependence on central infra, it increases edge-compute and maintenance overheads. Regulatory compliance adds another layer of complexity, with different frameworks (such as GDPR, India’s DPDP Act, and HIPAA) requiring alignment.

A Hybrid Future?

Despite the promise, none of the experts interviewed expect Big Tech to disappear. And rightly so, the consensus is on coexistence.

“The largest foundation models will still come from the big players because they have the resources. But decentralisation brings balance: startups and communities can fine-tune, localise, and deliver AI in ways Big Tech can’t,  closer to the edge, with more trust and sovereignty built in,” said Sripada.

Paul framed it as a relationship of scale and agility. “Big Tech will provide a stable foundation that acts as the backbone of the AI ecosystem. Startups and decentralised platforms will design sector-specific applications and services that make use of that backbone.”

For India, the significance extends beyond startups. Ganjoo called decentralisation an urgent matter of sovereignty: “Decentralised AI allows India to own the rails of intelligence, not just be a data supplier.”

As Sripada analogised, centralised GPTs may be the “Yahoo.com of 1997”,  impressive gateways, but only the beginning. An explosion may be seen once decentralisation allows niche models, agents, and services to bloom.

Startups then may no longer be challengers on the sidelines, but participate equally as architects of a democratic, transparent, and sovereign AI future.

The post How privacy and sovereignty could push the adoption of Decentralised AI appeared first on Analytics India Magazine.

Generated by RSStT. The copyright belongs to the original author.

Source