Google expands AI push in India with enterprise and security tools

At Google I/O Connect India 2026, the company unveiled enterprise AI, cybersecurity and education initiatives while expanding in-country AI infrastructure for businesses

Google is racing to position India as a proving ground for the next phase of artificial intelligence (AI), unveiling a slate of security protocols, on-premise computing options and education programmes on Tuesday.

 

The aim is to win over the country's enterprises, developers and startups as AI systems shift from answering questions to autonomously executing tasks.

 

At the Google I/O Connect India 2026 in Bengaluru, the company also highlighted that, based on third-party evaluations, the Google Play and Android ecosystem generated ₹5.3 trillion ($ 60 billion) in revenue for app publishers and the wider economy in India in 2025, growing by 28 per cent from 2024.

 

 

Manish Gupta, senior director for India and Asia-Pacific (APAC) at Google DeepMind, said the ultimate metric of AI progress isn't just model parameters, but also in the positive transformation it enables.

 

“India is championing this as it adopts AI across every tier of the economy — from local merchants to national health initiatives,” said Gupta.

 

He added, “In bringing our frontier AI, on-premise capabilities, and commitment to safety, we aim to accelerate this momentum, and look forward to the country’s AI learners, educators, builders, and innovators leading India’s AI ambition from the front.”

 

Google announced two initiatives aimed at expanding access to advanced AI skills in India, as the company seeks to help train the country's next generation of AI researchers, developers and educators.

 

One of the programmes brings Google DeepMind's AI Research Foundations curriculum — a free, 56-hour course designed to teach participants how to build and fine-tune large language models (LLMs) and conduct advanced AI research. Those who complete the programme can earn Google Cloud skill badges and professional certificates.

 

The programme will be available through Google's Skills platform and expanded through partnerships with IT body Nasscom and the Indian Institute of Science (IISc), Bangalore. It would later reach institutions across the country.

 

Separately, AVPN, a network of social investors, will work with local partners to deliver the curriculum to learners through the Google.org APAC AI Opportunity Fund.

 

Google DeepMind also launched ATL Saathi, a Gemini-powered assistant for teachers developed with the Atal Innovation Mission. The web-based tool will be introduced in 100 schools this year, with plans to eventually reach 10,000 schools covered by the Atal Tinkering Labs programme.

 

Google said researchers at AIIMS Delhi are using its open-source MedGemma AI models to develop India-specific tools for leprosy and sexual and reproductive health. The resulting models would be made available to the Indian developer ecosystem.

 

Separately, the company expanded Gemini Live to support 25 Indian languages and dialects, including Sanskrit, Bhojpuri and Maithili.

 

Google said it will make its Gemini 3.5 Flash model available to Indian businesses with in-country machine learning processing, expanding its push on data sovereignty.

 

The company is also enabling enterprises, including regulated industries and government agencies, to run Gemini on Google Distributed Cloud from data centres in India. It is allowing AI applications to operate entirely within their own infrastructure and without relying on the public internet.

 

Asked how Google DeepMind differentiates itself from rivals such as OpenAI, Anthropic, Meta and Microsoft in winning Indian developers and enterprises, Seshu Ajjarapu, senior director of applied AI at Google DeepMind, said the company's advantage lies in its end-to-end AI stack.

 

“Nothing separates us than our full-stack approach,” said Ajjarapu.

 

Ajjarapu said Google builds across the AI stack — from its infrastructure, including tensor processing units (TPUs), to foundation models, open models such as Gemma, domain-specific models, including MedGemma, and enterprise AI tools.

 

He said the breadth of Google's model portfolio, combined with products used by billions of people, allows the company to test and refine its AI models at scale before deploying them to enterprise customers.

 

Google is also expanding access to Sec-Gemini v3, its AI-powered cybersecurity agent, to selected government and enterprise testers, including Flipkart, as it seeks to address security risks posed by agentic AI. The company is also open-sourcing CAPSEM, a secure runtime environment for AI agents, and said it is working on industry standards for secure AI-driven transactions. Separately, Google announced research partnerships with IIT Delhi and IIT Madras focused on agentic AI safety and threat detection.

 

Google cited third-party research showing strong adoption of its AI tools among Indian developers and said it will expand its Google Play Academy curriculum to 10,000 developers through partnerships with the governments of Rajasthan, Odisha and Madhya Pradesh. At Google I/O Connect India, about 1,500 developers attended the event, while startups including Adya.AI, VideoSDK, Sivi, Superjoin, Knit, PolicyBazaar, Emergent and RedBus showcased applications built using Google's AI models.