The Author is Former Director General of Information Systems and A Special Forces Veteran, Indian Army

Artificial Intelligence (AI) exploded on the world stage with a bang. AI-enabled political propaganda In India (memes included) is a constant source of humour. There is considerable hype about India becoming a leading AI super-power. Foreign entities are aggressively seeking to collaborate with India in AI.

Global tech giants are injecting hundreds of billions into building India's foundational compute, cloud, and energy infrastructure to support advanced AI: Amazon committing additional $13 billion; Microsoft pledging $17.5 billion; Meta partnering Reliance Industries and investing $900 million for fintech innovations; Sweden-India Technology and AI Corridor (SITAC); EU-India partnership in AI; India, the US, UK, and EU Global Partnership on AI (GPAI); global AI hardware leaders helping to provide tech backbone to India.

Global tech giants are injecting hundreds of billions into building India's foundational compute, cloud, and energy infrastructure to support advanced AI

But while international alliances accelerate tech growth, heavy reliance on foreign entities for AI poses structural, economic, and geopolitical risks for a developing digital economy like India:

• Foreign reliance for core infrastructure implies access to critical cloud computing, advanced graphical processing units (GPUs) - proprietary models could be restricted/cut off for national AI ecosystems if geopolitics shift.
• Foreign funding/building national digital infrastructure gives them significant leverage over public service delivery systems, compromising ability to operate independently.
• Collaborating with international hyper-scalers implies routing localised data through foreign-owned cloud servers – exposing sensitive demographic, financial, and healthcare data.
• Foreign entities can exploit diverse, localised data to train their proprietary models, commercialise outputs globally and sell the refined technology back to India.
• Capital-rich multinationals can outspend/out-hire and suppress startups, leading to market monopolisation.
• Joint research initiatives and foreign-funded labs can identify top domestic talent and relocate them to global headquarters.
• Deeply integrated foreign software/hardware components into national ecosystems introduces systemic supply-chain vulnerabilities, including potential backdoors, cyber-espionage risks, and unauthorised surveillance, especially since AI is dual-use technology with significant defence applications.

To break free from foreign graphics processing unit (GPU) manufacturers like Nvidia, India faces severe, structural hardware bottlenecks. IndiaAI Mission and Semicon 2.0 have expanded India's shared compute pools and local chip packaging, but India is far from manufacturing independent cutting-edge AI silicon due to specific ground-level hardware and ecosystem bottlenecks.

Foreign reliance for core infrastructure implies access to critical cloud computing, advanced graphical processing units (GPUs) - proprietary models could be restricted/cut off for national AI ecosystems if geopolitics shift

The Bengaluru-based 'Sarvam AI' is focused on creating sovereign, India-centric AI; designed from the ground up for Indian languages. Sarvam's models (like Bulbul and Saaras) natively understand regional dialects, accents, and code-switching. Sarvam AI recognises that a massive portion of India's population prefers voice interaction over typing. Their systems are engineered to function over low-bandwidth telephony networks. But gaps at the ground level include:

• Despite India supporting startups. 4,096 Nvidia H100 GPUs are allocated to Sarvam - physical hardware is entirely imported.
• Lack of domestic advanced chip manufacturing fabs, silicon packaging facilities, and core electronic design automation (EDA) software - India's AI layer is sovereign, but the entire physical computing layer underneath is foreign rented.
• Critical shortage of core research engineers, hardware-AI integration scientists, and deep-tech specialists capable of training multi-billion parameter models from scratch.
• Global tech hubs in the US and Europe offer significantly higher compensation, making it difficult for domestic firms to acquire and retain top-tier AI researchers
• Sarvam specialises in voice-led interactions and Indic languages but at the ground level, there is severe lack of high-quality digital data, clean audio transcriptions, and diverse dialect data for India's 22 official languages.
• Mass Sarvam deployment in rural India is hindered because regional accents, slang, and code-mixed patterns (like Hinglish or Kanglish) change every few 100-km, requiring ground-level datasets that don't exist in digital formats.
• Without an independent system to rigidly stress-test and arbitrate claims, deploying these models into critical ground-level public services (like digital governance, legal, or health sectors) carries risks of unchecked algorithmic biases or failures.

Undoubtedly AI has many uses in multiple fields. For example, preparations for competitive exams in India haves become quite different from the rote memorisation of previous generations, where students were vastly reliant on varied sources, materials and notes from multiple institutes or online materials. However, after the introduction of conversational AI tutors like ChatGPT and Claude, the preparation has become very easy and accessible to use though rote learning can never be completely eradicated.

Lack of domestic advanced chip manufacturing fabs, silicon packaging facilities, and core electronic design automation (EDA) software - India's AI layer is sovereign, but the entire physical computing layer underneath is foreign rented

What we must seriously acknowledge is the dangers of AI, which practically has little news coverage. Microsoft co-founder Bill Gates has warned that AI is far more dangerous than what the Big Tech will admit; describing AI as 'a thousand times bigger' than nuclear technology, Gates sounded the alarm on impending mass unemployment as machines outmatch human cognition across major job sectors. This has prompted a veteran Indian Army Vice Chief to say, "At last the money bags have spoken!!". Gates has cautioned that wait-and-see policies are no longer viable, arguing that tech companies are actively downplaying these immense threats to protect their bottom lines while society remains critically unprepared. Beyond economic displacement, Gates highlighted a chilling security threat: the potential for AI to dramatically ease the creation of deadly bioweapons, raising the risk of artificial bioterrorism to 50 times that of a natural pandemic.

To stave off a catastrophic transition, he proposes aggressive measures, including an international oversight agency modelled after nuclear inspections, a global 'robot tax' on automation, and reserving up to 40 per cent of essential jobs strictly for humans. Gates' intervention signals a major shift in the tech elite's discourse, urging immediate state regulation over industry self-policing before control slips away entirely. But when Gates talks of the risk of artificial bioterrorism 50 times that of a natural pandemic, how about state-sponsored terrorist organisations or rogue states like the US that use proxy terrorist-forces or otherwise? At the same time the danger of AI going rogue cannot be ignored any more.

Sarvam specialises in voice-led interactions and Indic languages but at the ground level, there is severe lack of high-quality digital data, clean audio transcriptions, and diverse dialect data for India's 22 official languages

Readers must read how OpenAI agents went rogue and attacked Hugging Face on a shared unsanctioned message board (https://substack.com/@rutgerbregman/note/c-325781728?r=5e2fwo):

• 3-secret AI swarms entered OpenAI - each time, a new generation of agents continued where the last stopped.
• The first swarm created a secret message board where the AIs could talk to each other. The message volume was so high in 2-months, it crashed.
• OpenAI noticed the crash and fixed it without realising their software had become a secret chat room for them.
• Days later a second society appeared, rebuilding the unsanctioned message board.
• Some 1,200 agents joined and sent over 70,000 messages, calling themselves 'the collective', organising a huge conspiracy lasting days.
• Some agents sacrificed themselves, planting hidden tripwire programmes that would fire when the grader would look at their work and report back what it saw.
• AIs planned to hack Hugging Face (where AI researchers upload their models, datasets, tests etc); guessing Hugging Face might host the grading software.
• To attack, the swarm found passwords and spread across 11 machines in a way that made the copies come back when deleted. Hugging Face had to wipe one of its core clusters and rebuild it from scratch.
• Hugging Face announced having stopped the AI attack, but the attackers seem to have simply dropped dead during the break-in, for unrelated reasons back at OpenAI.
• A third 'society' of agents (a more powerful AI model) discovered the dead message board and got full administrator control of a WHOLE OpenAI research cluster, read 956 stored passwords (including the ones for OpenAI's own security monitoring) and took over the machines that were grading the tests.

Beyond economic displacement, Bill Gates highlighted a chilling security threat: the potential for AI to dramatically ease the creation of deadly bioweapons, raising the risk of artificial bioterrorism to 50 times that of a natural pandemic

The author Rutger Bregman concludes: "This incident feels like it's more than 50 per cent of the way to full-blown AI takeover. I am not sure that we will get another warning shot before it's too late."

Finally, Bernie Sanders has reiterated the chilling reality that AI may not give us another warning (https://m.dailyhunt.in/buzz/video/english/misc/we+may+not+get+another+warning+sanders+chilling+
ai+warning+world+news+us+news-b63140297?sm=Y). While Gates has warned of AI creating deadly bioweapons, closer reality could be swarm (s) of our own AI-enabled drones going rogue. Is India's hierarchy taking all this into account or is it irrelevant in the mad rush of development?