Every technological revolution has its defining moment. The internet connected the world, smartphones placed computing in every pocket, and generative AI taught machines to communicate with remarkable fluency. Now another transformation is underway. A new generation of “agentic” AI can plan, reason and act. Given an objective, these systems can coordinate multiple applications, book travel, write software or manage business workflows with minimal supervision. The chatbot waits for instructions; the AI agent gets to work. This seemingly subtle shift from answering questions to taking action may prove to be one of the most consequential technological changes of our time.
Even more remarkable is that much of this capability is available at little or no cost. Basic AI assistants are free, while advanced models are offered through subscriptions or usage-based pricing. Providers typically charge by “tokens” which are small units of text that measure both the prompts’ users submit and the responses they receive. A simple query costs very little, but analysing lengthy documents or running autonomous AI agents continuously can quickly become expensive.
The global AI race is becoming equally significant. American companies such as OpenAI, Google and Anthropic dominate premium cloud-based AI, while Chinese firms including DeepSeek and Alibaba have disrupted the market with capable models offered at a fraction of the cost, many available as open downloads. For consumers, this means more choice and lower prices. For the Government, it raises a more strategic question: whose technology and whose values will underpin the digital infrastructure of the future?
Another divide concerns how AI itself is built. Closed models are proprietary services whose inner workings remain secret. Open models can be downloaded, examined and run on an organisation’s own computers. Open AI models offer greater transparency, lower costs and technological independence, but also increases the risk of misuse. Closed models maybe more secure but concentrate power in a handful of companies and leave users dependent on their policies and pricing. The choice is ultimately between openness and autonomy on one side, and convenience and managed security on the other.
As AI becomes woven into everyday life, privacy deserves far greater attention. Every prompt entered into an AI system may be stored and, depending on the provider, used to improve future models. Medical records, legal documents, business strategies and trade secrets can become vulnerable through careless use or security breaches. The safest rule is simple: treat an AI chat like a postcard, not a sealed envelope. Never upload anything you would not want others to read. For businesses and governments, a single careless upload could expose years of research, intellectual property or sensitive national information.
AI also has clear limits. It excels at repetitive, structured work such as summarising reports, translating documents, analysing data and drafting routine correspondence. But decisions requiring judgment, empathy and accountability should remain firmly in human hands. Medicine, justice, education, military operations, hiring and the care of children or the elderly are areas where AI should advise and not decide. Responsibility cannot be delegated to an algorithm.
The broader consequences are profound. AI will eliminate some jobs while reshaping many others, demanding continuous reskilling throughout working lives. Older workers may find decades of expertise suddenly devalued by technologies that perform routine knowledge work in seconds. At the same time, as conversations with machines increasingly replace conversations with people, convenience may come at the cost of genuine human interaction.
Perhaps the greatest long-term risk is cognitive dependence. If machines increasingly perform our writing, memory, arithmetic and reasoning, our own mental abilities may gradually weaken through disuse. Children who never struggle to solve problems may never develop resilience or critical thinking. AI should be used to augment human intelligence, not replace it.
There is also an environmental price. The data centres powering AI consume enormous quantities of electricity and water for cooling. Together they used an estimated 448 terawatt-hours of electricity in 2025, more than the annual consumption of many countries and demand is expected to rise sharply. Every seemingly effortless AI response depends on vast physical infrastructure whose environmental costs remain largely invisible to users.
Finally, AI is becoming an issue of national security. If our phones, hospitals, vehicles, financial systems and public services depend on AI controlled from abroad, technological dependence becomes strategic dependence. In a crisis, foreign-controlled AI could become a channel for surveillance, coercion or disruption. Cheap imported AI may seem an irresistible bargain today, but it could quietly mortgage our technological sovereignty tomorrow.
None of this argues against artificial intelligence. Few technologies offer greater potential to transform healthcare, education, scientific discovery, governance and economic productivity. But powerful technologies demand equally strong safeguards. We must establish sensible regulation, companies must uphold privacy and transparency, and we must invest in home-grown AI capabilities rather than becoming passive consumers of foreign technology.
Artificial intelligence is likely to become as indispensable as electricity or the internet. Handled thoughtfully, it could become humanity’s greatest intellectual ally. Handled carelessly, it may prove to be the cleverest mistake we ever make.
(Brig Anil John Pereira, SM (Retd) is a veteran from Goa, who served the nation with distinction for 32 years).
