At a moment when artificial intelligence is shifting from answering questions to acting on behalf of human institutions, Google has chosen India as its most ambitious laboratory. Across a single week's announcements, the company wove together tools for classrooms, hospitals, government agencies, and startups — each designed to meet India where it is, in its languages, its regulations, and its aspirations. The gesture is both commercial and philosophical: a wager that the next chapter of AI's story will be written not in one language or one data center, but in twenty-five languages and across a
Google expands AI infrastructure in India with teacher app, localized cloud and security tools
AI moves from answering queries to securely executing tasks
Mark
Why does Google care so much about India specifically? There are AI users everywhere.
Mimi
India has 1.4 billion people and a massive developer ecosystem, but it's also a place where Google has to prove it can build AI that works across dozens of languages, respects strict data residency laws, and serves everything from small merchants to national health systems. It's not just a market—it's a complexity test.
Mark
This "agentic AI" phrase keeps appearing. What does that actually mean in practice?
Mimi
Right now, AI mostly answers your questions. Agentic AI would actually do things—execute transactions, investigate security incidents, make decisions—on your behalf. That's more powerful but also more dangerous if something goes wrong, which is why Google is emphasizing security so heavily.
Mark
The data localization piece seems important. Why would an Indian bank or hospital care if their data stays in India versus going to a Google server elsewhere?
Mimi
Regulation, partly. India has strict rules about where certain data can be stored. But there's also trust—if your health records or financial data never leave the country, you have more control over it. Google is saying: build your AI systems entirely within your own borders, on our infrastructure.
Mark
What's the play with all these Indian languages—Sanskrit, Bhojpuri? Those aren't exactly high-volume commercial languages.
Mimi
It signals that Google sees India as a place where AI needs to serve everyone, not just English speakers in cities. If you can build AI that works for a farmer in Bihar or a student in a village school, you've solved a harder problem than building it for San Francisco.
Mark
The security tools—Sec-Gemini, CAPSEM—feel like they're preparing for something.
Mimi
They are. As AI systems become more autonomous and handle real transactions or sensitive data, the attack surface grows. Google is essentially saying: we're not going to deploy this technology without the guardrails already in place. It's defensive positioning, but it's also necessary.
The Pulse
- Google's pivot from AI as a query tool to AI as an autonomous executor is arriving in India first, raising urgent questions about safety, data sovereignty, and who controls the infrastructure of daily life.
- With Rs 5.3 lakh crore paid to Indian developers in a single year and 91% reporting revenue gains from AI tools, the economic stakes of this expansion are already reshaping the developer ecosystem.
- Regulated sectors — healthcare, finance, government — are being courted with data localization promises, as Gemini models can now run entirely within Indian borders on isolated infrastructure, directly addressing fears of sensitive data crossing jurisdictions.
- Security is the load-bearing wall of the entire announcement: Sec-Gemini v3, the CAPSEM open-source framework, and IIT partnerships on 'Guardian Agents' all signal that Google knows autonomous AI without guardrails is a liability, not a product.
- The inclusion of Sanskrit, Bhojpuri, and Maithili in Gemini Live, alongside MedGemma models targeting leprosy and reproductive health at AIIMS Delhi, suggests Google is attempting something rare — infrastructure that bends toward local reality rather than demanding local reality bend toward it.
At a moment when artificial intelligence is shifting from answering questions to acting on behalf of human institutions, Google has chosen India as its most ambitious laboratory. Across a single week's announcements, the company wove together tools for classrooms, hospitals, government agencies, and startups — each designed to meet India where it is, in its languages, its regulations, and its aspirations. The gesture is both commercial and philosophical: a wager that the next chapter of AI's story will be written not in one language or one data center, but in twenty-five languages and across a subcontinent of extraordinary complexity.
Google has placed a large and deliberate bet on India as the proving ground for agentic AI — systems that don't merely respond to questions but carry out tasks on behalf of users and institutions. At its India I/O Connect event, attended by more than 1,500 developers, the company unveiled a constellation of tools designed to embed AI into enterprises, hospitals, classrooms, and government agencies, with security and local relevance as the twin organizing principles.
The company's existing presence in India is already formidable. In 2024, Google distributed over Rs 5.3 lakh crore to developers on its Android and Play platforms — a 28 percent increase year-on-year — and nine in ten developers who used its AI tools reported a direct boost to their earnings. Three-quarters of Indian adults expressed interest in generating code simply by describing what they wanted in plain language, a sign that AI's reach is extending well beyond technical communities.
The new offerings reflect a deliberate effort to meet India's specific conditions. Gemini Live now supports 25 Indian languages and dialects, including Sanskrit, Bhojpuri, and Maithili. For sectors where data cannot leave Indian borders, Google announced that Gemini models can now run entirely within domestic data centers on infrastructure isolated from the public internet. At AIIMS Delhi, researchers are using Google's MedGemma models to build AI systems tailored to Indian health challenges, including leprosy and sexual and reproductive health.
Security ran as a central thread through every announcement. Sec-Gemini v3, a specialized cybersecurity agent, is being made available to trusted government and enterprise partners including Flipkart, capable of analyzing threats at machine speed. Google is also open-sourcing CAPSEM, a framework that keeps AI infrastructure safe even when individual models are manipulated by malicious prompts. Researchers at IIT Delhi and IIT Madras are collaborating with Google on 'Guardian Agents' — AI systems designed to detect and contain institutional risks before they escalate.
Country Manager Preeti Lobana described the moment as a structural shift: as AI moves from answering queries to executing tasks, she said, the priority is building the infrastructure and guardrails the ecosystem needs to scale safely. For Google, India is simultaneously its largest emerging market and its most complex test case — a place where the ambitions of the agentic era must contend with twenty-five languages, strict regulatory environments, and the full spectrum of human need.
Google is betting heavily that India will be the proving ground for a new kind of artificial intelligence—one that doesn't just answer questions but actually executes tasks on behalf of users and organizations. On Tuesday, the company announced a suite of tools and partnerships designed to embed AI across Indian enterprises, startups, classrooms, and hospitals, with a particular emphasis on security and local relevance.
The scale of Google's existing footprint in India is already substantial. Last year, the company distributed over Rs 5.3 lakh crore to developers building apps and services on its Android and Play platforms—a 28 percent jump from the previous year. Among developers surveyed who had used Google's AI tools, 91 percent said those tools had directly or indirectly helped them earn more money. Three-quarters of Indian adults expressed interest in using AI to write code simply by describing what they wanted in plain language, a practice known as "vibe coding."
But Google's ambitions extend far beyond app developers. At its India I/O Connect event, which drew over 1,500 developers, the company rolled out a series of localized offerings. Gemini Live, its conversational AI assistant, now works in 25 Indian languages and dialects—including Sanskrit, Bhojpuri, and Maithili. For regulated industries and government agencies concerned about data leaving Indian borders, Google announced that enterprises can now run Gemini models entirely within Indian data centers using Google Distributed Cloud, with supporting infrastructure fully isolated from the public internet. This addresses a critical concern for sectors like healthcare and finance, where data residency requirements are strict.
In healthcare specifically, Google made its MedGemma models available to researchers at AIIMS Delhi, who are using the technology to develop AI systems tailored to Indian health challenges—leprosy and sexual and reproductive health among them. The company is also open-sourcing CAPSEM, a security framework that protects AI systems even if they are compromised or tricked by malicious prompts, ensuring the wider infrastructure remains safe.
Security emerged as a central theme across Google's announcements. The company is bringing Sec-Gemini v3, a specialized cybersecurity agent, to trusted government and enterprise testers including Flipkart. This tool can analyze complex security data and investigate incidents at machine speed, helping organizations defend against threats that might arise as AI systems become more autonomous. Google is also introducing industry standards like Device Bound Session Credentials, which renders stolen session cookies useless to attackers, and new protocols for secure, low-value financial transactions conducted by AI agents.
Researchers at IIT Delhi and IIT Madras are collaborating with Google on agentic safety research, developing what the company calls "Guardian Agents"—AI systems designed to detect threats early and mitigate institutional risks. This academic partnership signals that Google sees India's technical community as essential to solving the safety challenges that come with more autonomous AI systems.
Preeti Lobana, Google's Country Manager for India, framed the announcements as part of a broader shift. "As we move into the agentic era, where AI moves from answering queries to securely executing tasks, our focus is on providing the underlying infrastructure and guardrails the ecosystem needs to scale safely," she said. The company is also expanding access to its AI Research Foundations program, a free 56-hour curriculum for building and fine-tuning large language models, in partnership with Nasscom, IISc Bangalore, and AVPN, a network of social investors.
Indian startups showcased their own AI solutions at the event, including companies like Adya.AI, VideoSDK, PolicyBazaar, and RedBus. For Google, India represents both a massive market of potential users and a testing ground for how AI infrastructure can be built to serve diverse languages, regulatory environments, and economic sectors simultaneously.
Notable Quotes
As we move into the agentic era, where AI moves from answering queries to securely executing tasks, our focus is on providing the underlying infrastructure and guardrails the ecosystem needs to scale safely.— Preeti Lobana, Country Manager, Google India
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.— Dr Manish Gupta, Senior Director for India and APAC, Google DeepMind
Want the full story?
Read the original at The Indian Express ↗
Technology