AI Strategy - India

 

AI Strategy - India

R Kannan

For decades, the global narrative about India’s technology sector was simple: it was the "back-office of the world." Indian engineers built the software that kept global banks running, airlines flying, and retailers selling. We were the masters of maintenance and the architects of efficient service. But today, a profound transformation is underway. India is no longer just maintaining the world’s digital infrastructure; it is actively building the next generation of it.

As we look toward 2026 and beyond, India is positioning itself to be more than just a participant in the AI revolution—it intends to be a leader. The ambition is not merely to "adopt" AI but to create an AI ecosystem that is uniquely Indian, highly efficient, and deeply inclusive.

From Service to Sovereign Innovation

The shift started with a realization: relying entirely on foreign-made, "black-box" AI models is neither sustainable nor sovereign. With massive investments flowing into India from global giants like Google and Microsoft, and domestic powerhouses like Tata, Reliance, and Adani doubling down on AI, the capital is there. But capital is not the solution; strategy is.

The Indian government, in collaboration with industry leaders, has begun to craft a roadmap that acknowledges a simple truth: we cannot just copy the Silicon Valley model of "bigger is better." We have a unique set of constraints—energy availability, data diversity, and the need for extreme cost-efficiency. Our strategy must be "AI-native."

This starts with infrastructure. We are moving toward a "Compute-as-a-Service" model. By providing subsidized GPU access, we ensure that an AI startup in a tier-two city has the same mathematical firepower as a legacy firm in a major metropolis. We are also mandating "Green Data Centre" policies. India cannot afford the massive energy footprint of Western-style data centres. By pushing for liquid cooling and renewable energy integration, we are not just building AI; we are building sustainable AI.

The SLM Revolution: Efficiency over Size

The most exciting aspect of India’s approach is the strategic pivot to Small Language Models (SLMs). While the world remains obsessed with building ever-larger Large Language Models (LLMs)—which require billions of dollars in electricity and processing power—India is championing the "smart-sizing" of AI.

Leaders like Nandan Nilekani have correctly identified that for a nation of 1.4 billion people, efficiency is the ultimate feature. An SLM trained on high-quality, domain-specific Indian data can perform better in local contexts than a massive, generalized model trained on Western internet data. By focusing on SLMs, we lower the cost, reduce energy consumption, and make AI deployable on mobile devices. This is the "Democratization of AI"—bringing intelligence to the fingertips of the farmer, the shopkeeper, and the student.

To fuel this, we are unlocking the "AIKosh"—our national data repository. This is not just a digital warehouse; it is a strategic asset. By curating non-personal data from government ministries, we are creating datasets that reflect the nuance, the dialects, and the complex reality of life in India. In the world of AI, data is the new oil, and India is refining it to create high-octane fuel for its own indigenous models.

Transforming the Human Capital

The most critical component of this strategy remains our people. India produces more engineers every year than almost any other nation. However, the challenge is not quantity; it is relevance.

We are currently witnessing a massive, state-backed effort to "re-tool" the workforce. The IT services giants—TCS, Infosys, Wipro, and HCL—are not just observing the AI wave; they are training their vast armies of employees to ride it. By incentivizing the private sector to pivot from "Legacy IT" to "AI-Native" roles, we are protecting our most valuable asset: our workforce.

Furthermore, we are rethinking education. We are moving away from theoretical coding toward "Applied AI." By integrating real-world project-based learning into engineering curricula and establishing innovation labs in regional colleges, we are ensuring that the talent pool is not concentrated solely in the metros. This decentralization of talent is essential to prevent the social disparities that often accompany rapid technological change.

The Governance of Trust

As India scales its AI infrastructure, it is also setting a global example for governance. The world is grappling with the ethical dilemmas of AI—bias, deepfakes, and job displacement. India’s approach, characterized by a commitment to "AI for All," prioritizes trust and transparency.

Through the development of Explainable AI (XAI) standards, we are ensuring that when an AI system makes a decision—whether it’s approving a loan or diagnosing a medical condition—that decision can be audited. This is crucial for maintaining public trust. We are also building "regulatory sandboxes." These controlled environments allow startups to innovate, test, and fail safely without the burden of full-scale regulation, accelerating the pace of breakthrough inventions.

Moreover, by actively participating in global governance forums like the Global Partnership on AI (GPAI), India is ensuring that the "Global South" has a seat at the table. We are proving that you do not need to choose between rapid economic growth and ethical development.

The Path Forward

Is this vision easy to execute? Certainly not. We face significant hurdles. Building the semiconductor fabrication units (ATMP) required to reduce our reliance on imported silicon is a decade-long project. Coordinating the "AI-Mandate" across government, where every major tender must demonstrate an AI efficiency boost, requires a cultural shift in bureaucracy. And ensuring the "Reverse Brain Drain"—bringing our best research scientists back to India—requires a competitive ecosystem of salaries, research freedom, and prestige.

However, the foundation is set. We have the Digital Public Infrastructure (DPI) legacy of Aadhaar and UPI, which has already taught us how to scale technology to a billion people. We have a private sector that is eager to invest, and a government that is creating the policy "rails" for this train to run on.

The strategy is comprehensive. From fostering open-source indigenous frameworks to creating sector-specific Centres of Excellence in agriculture and healthcare, every piece of the puzzle is designed to create a self-sustaining loop of innovation.

In the global AI race, many nations are currently focused on the "how"—how to build a bigger model, how to get more GPUs, how to control the market. India’s focus is different. We are focused on the "who" and the "what." Who is this for? It is for the billions who have been underserved. What is it for? It is to solve the complex problems of healthcare, agriculture, and education that standard Western models often overlook.

By combining sovereign infrastructure, efficient model development, a massive upskilled workforce, and an ethical regulatory framework, India is not just catching up. It is crafting a blueprint for the future. The world once looked at India as a place where the world’s problems were outsourced for solution. Now, the world is looking to India to see how, in the age of AI, the human potential of a nation can be unlocked at a scale never before imagined. The AI era has arrived, and for the first time in history, India is leading the charge, not just as a provider of services, but as the architect of the future.

For Detailed report . Contact : rajakannan@rediffmail.com

 

Beyond the Hype: Why You Need to Know the Three Faces of AI

 

Beyond the Hype: Why You Need to Know the Three Faces of AI

R Kannan

The world today is buzzing with one word: "AI." You hear it in boardrooms, read about it in newspapers, and see it in every piece of software you use. But there is a massive amount of confusion. People talk about AI as if it is a single, magical box that solves every problem.

 

This is a dangerous misconception. Using the term "AI" to describe everything from a simple spam filter to a complex autonomous agent is like calling a bicycle, a fighter jet, and a cruise ship all just "transportation." They all help you move, but they serve completely different purposes, require different skills to operate, and carry very different risks.

If we want to build a future where technology actually helps us—rather than just adding noise—we need to stop looking at AI as one giant concept. We need to understand that we are living in a three-stage evolution: the Analyst, the Creator, and the Doer.

The Analyst: Traditional AI

Let’s start with the "Analyst." This is what we have been using for decades. When you see a bank block a suspicious credit card transaction, or when your streaming service recommends a movie you actually enjoy, you are looking at Traditional AI.

Its core philosophy is classification and prediction. It is designed to look at a pile of data and say, "This is what that is," or "This is what will likely happen next."

Why is this useful? Because it is incredibly precise. It doesn’t "hallucinate" or make up facts. It works on strict rules and patterns. If you need to detect fraud, optimize a logistics route to save fuel, or spot a tumour in an X-ray, you don’t want a machine to be "creative." You want it to be accurate.

Traditional AI is the backbone of efficiency. It is the workhorse of the digital world. It doesn’t need to be "smart" in a human way; it just needs to be better at recognizing patterns than a human can be. The value here is reliability. When you rely on this, you are betting on the stability of the math.

The Creator: Generative AI

Then, we have the "Creator." This is the technology that exploded onto the scene recently with tools like ChatGPT, Gemini, and Midjourney.

Generative AI shifted the goalpost entirely. Its purpose is not to predict the past or classify data; its purpose is to synthesize and create. It learns the patterns of human language, code, or art, and then it produces new, original content based on those patterns.

This is where things get exciting—and tricky. Generative AI allows for a massive leap in speed and creativity. Suddenly, you can draft marketing emails in seconds, write complex code snippets, or create illustrations for a presentation without having to be a professional designer. In education, it can act as a personal tutor that explains a concept in five different ways until a student understands it.

But here is the catch: The Creator is not an Analyst. It can be wrong. It can sound incredibly confident while saying something completely incorrect, because its goal is to be plausible, not necessarily true. This is what experts call "hallucination."

If you use Generative AI as your sole source of truth, you will eventually fail. But if you use it as a brainstorming partner, a first-draft writer, or a tool to help you synthesize information, it provides a level of leverage that was impossible just a few years ago.

The Doer: AI Agents

Now we arrive at the most important frontier: the "Doer," or AI Agents.

If Traditional AI is the Analyst and Generative AI is the Creator, AI Agents are the Employees. They are systems capable of planning, using tools, and executing complex, multi-step goals.

Think about the difference. You can ask a chatbot (Generative AI) to "write an email to my sales lead." But you still have to copy that text, open your email app, find the lead’s address, paste the text, and hit send.

An AI Agent changes the game. You simply say, "Research this lead and reach out to them." The Agent will search the web for the lead’s company news, draft the email, check your CRM to see if they are already in the system, and then send the message. It doesn't just give you the answer; it does the work.

Agents work in a loop: Plan, Act, Observe. If they encounter an error—say, the website they need to check is down—they don't just stop. They think, "The site is down," and they try a different approach. They can use external tools, APIs, and software applications just like a human would.

This is the future of labour automation. Agents are the "missing link" that connects the intelligence of Generative AI with the utility of software systems. They are ideal for complex workflows like managing supply chains, conducting deep research, or running IT helpdesk support.

How to Think About Your Own Future

So, why does this distinction matter for us ?

Because most leaders and individuals are currently making the same mistake: they are trying to solve every problem with a hammer, even when they need a screwdriver.

If you are trying to automate a boring, repetitive task that requires 100% accuracy, do not look for a chatbot. You need Traditional AI. You need an "Analyst" that works on logic and numbers.

If you are stuck on a blank page, if your marketing team is burnt out, or if you need to understand a massive volume of documents quickly, you need a "Creator." You need Generative AI to boost your speed and break through your creative block.

And if you are tired of clicking buttons, copy-pasting data between apps, and managing manual, multi-step workflows, you need a "Doer." You need AI Agents to handle the heavy lifting of execution.

The Human Element

There is a final, crucial point to make about this evolution. As these tools become more powerful, the value of the human being actually increases, not decreases.

AI, in all its forms, is essentially a tool. The "Analyst" provides the insight. The "Creator" provides the options. The "Doer" provides the output. But the human? The human provides the judgment.

A machine can generate a hundred marketing slogans, but it cannot tell you which one aligns with your company’s soul. A machine can research a lead, but it cannot determine the right tone to use in a negotiation. A machine can optimize a route, but it cannot decide if that route is the most ethical path for your business.

We are moving away from an era where we needed to be "manual labourers" of information—typing, copying, pasting, classifying—and moving into an era where we are the "architects" of our own work.

The technology is getting better at answering, creating, and doing. Our job is to get better at asking, guiding, and deciding.

Don’t be intimidated by the pace of change. Stop worrying about "AI" as if it were a single, incoming tide that will wash everything away. Instead, learn to identify the tools. Build your team of "Analysts," "Creators," and "Doers" using the best technology available.

The future doesn't belong to those who fear the machine. It belongs to those who know exactly which machine to turn on, why they turned it on, and—most importantly—when to leave it to the humans.

For the detailed report : Contact rajakannnan@rediffmail.com

 

A Manifesto for Labor in the Age of Artificial Intelligence

 

A Manifesto for Labor in the Age of Artificial Intelligence

R Kannan

As Labor Day 2026 arrives, the global workforce stands at a juncture as pivotal as the Industrial Revolution. Yet, unlike the steam engine, which replaced muscle with machinery, artificial intelligence (AI) is beginning to substitute, augment, and redefine the very cognitive processes that have defined human labour for centuries.

For the past decade, we have debated whether AI would lead to a "jobless future." Today, as we analyse the early evidence—including the sobering insights from the 2026 Joint ILO-World Bank working papers—we see that the reality is more nuanced, and perhaps more urgent. The question is no longer if AI will change the nature of work, but *how* we can govern that change to prevent the widening of global inequalities.

 

The Great Divergence: Exposure vs. Readiness

Recent data from the World Bank and the International Labour Organization (ILO) underscores a critical reality: AI’s impact is inherently uneven. In advanced economies, where digital infrastructure is ubiquitous, AI exposure is high—reaching up to 30–32% of employment. Here, the challenge is managing the transition for clerical and professional roles that are susceptible to automation.

Conversely, developing economies face a different, perhaps more insidious, risk. While their overall exposure to AI automation is lower, they suffer from a "readiness gap." As noted in the 2026 *Digital Progress and Trends Report*, the lack of robust digital connectivity and AI-ready infrastructure threatens to trap these nations in low-productivity cycles. If these countries cannot leapfrog into AI-enabled service delivery, they risk losing the traditional "escalator" to development: the expansion of manufacturing and service-sector jobs that previously pulled millions out of poverty.

The result is a looming "Great Divergence." If we leave market forces entirely to their own devices, we risk a world where the AI-dividend accrues disproportionately to capital-rich nations, while labour-rich developing nations struggle with stagnation.

Beyond Automation: The Augmentation Imperative

The fear of job displacement is palpable, but the IMF’s analysis of 2026 labour trends suggests a more complex dynamic: polarization. We are observing the emergence of a "skill premium" where workers who can leverage AI to augment their output see rising wages, while those in routine, non-complementary roles face wage suppression or displacement.

The goal for policymakers cannot be to stop the machine; it must be to change the machine’s objective function. Governments must move from a defensive stance—trying to protect obsolete jobs—to an offensive strategy of human-centric augmentation.

We must distinguish between AI that serves to replace human judgment and AI that serves to amplify it. Tax incentives should be restructured to reward firms that use AI to upskill their workforce, rather than those that simply use automation to trim headcount. This is not just a moral imperative; it is an economic one. As the World Bank’s 2026 Spring Meetings emphasized, "jobless growth" is a dead end. Sustainable development requires the active participation of the workforce in the value-creation process.

A Global Roadmap for Human-Centric AI

To navigate this transition, governments must adopt a comprehensive policy architecture. I propose a 12-point framework, built on the necessity of proactive governance:

 1. AI-Augmentation Incentives: Transition tax systems to prioritize "human-plus-AI" models. Corporations that retrain staff to work alongside AI should receive tax credits equivalent to capital investment incentives.

 2. Universal Lifelong Learning Accounts (ULLA):Education can no longer be a front-loaded, one-time investment. Governments should fund portable accounts, allowing workers to access modular, industry-certified training as market needs shift.

 3. Predictive Labor Market Intelligence: Using AI to govern AI, states should invest in predictive systems that identify, with 18-to-24-month lead times, which job roles are at risk, triggering proactive re-skilling pathways.

 4. Regulatory "Human-in-the-Loop" Standards: In high-stakes domains—healthcare, law, and financial advice—legislation must mandate human oversight, ensuring that AI provides decision-support rather than autonomous decision-making.

 5. Digital Public Infrastructure (DPI) Expansion: Governments must treat connectivity as a public utility. As India’s UPI model demonstrates, DPI lowers the cost of entry for small entrepreneurs, sparking mass-market job creation.

 6. Portable Social Security for the Gig Economy: The future of work is fragmented. We need a social safety net that follows the worker, not the workplace, covering health and retirement for gig and freelance contributors.

 7. Entrepreneurial Friction Reduction: Startups focused on "human-centric" technology—those that solve real-world problems in aging, education, and rural development—should face zero regulatory hurdles.

 8. Reskilling Mandates in Procurement: Public contracts should require that a percentage of the contract value be reinvested into local workforce development programs.

 9. Automation Levies: For high-profit, hyper-automated, labour-displacing processes, states should explore targeted levies. These funds must be ring-fenced exclusively for national reskilling initiatives.

 10. Curriculum-to-Industry Feedback Loops: National education councils must be redesigned to have industry leaders as permanent members, ensuring academic curricula are refreshed every 24 months.

 11. Collaborative AI Governance: Establish tripartite councils—government, industry, and academia—to set ethical and technical standards for AI deployment in the local economy.

 12. Inclusion for the "Last Mile": Prioritize digital literacy for rural and informal sectors to ensure that AI does not create a two-tiered economy of the "connected" and the "cut-off."

The Indian Laboratory: A Model for the Global South

India, with its vibrant demographic dividend and rapid digital maturation, stands as a critical microcosm for the world. The country’s path toward creating the millions of jobs required to eradicate poverty is no longer through mass assembly lines alone, but through a hybrid model of "High-Tech, High-Touch" development.

We see this already in the ten sectors of massive growth:

 The Green Transition: The shift to net-zero is perhaps the largest employment multiplier of the decade. From solar grid management to battery recycling, the "Green Collar" workforce is the future of sustainable labour.

 The Care Economy:  As the world ages, the "human touch" in nursing and elderly care is becoming an irreplaceable premium service. India is uniquely positioned to professionalize and scale this sector for domestic and global demand.

 The Creative Economy (AVGC):Animation, Visual Effects, Gaming, and Comics are not mere entertainment; they are the new frontier of digital manufacturing, leveraging India’s vast pool of artistic and technical talent.

The success of these sectors depends on integrating AI not as a competitor, but as a catalyst for efficiency. If India can successfully pilot this model—combining aggressive DPI expansion with massive, decentralized skill development—it will provide a template for the Global South to bypass the "middle-income trap" that AI threatens to worsen.

Conclusion

This Labor Day, we must resist the narrative of technological determinism. We are not passengers in a runaway train. We are the architects of the track.

The World Bank’s 2026 data serves as both a warning and a guide: technology will be an amplifier of existing trends. If we prioritize equity, it will accelerate progress. If we prioritize unfettered capital, it will accelerate inequality. The challenge of our time is to weave AI into the social fabric in a way that respects human dignity and expands the boundaries of what is possible for every worker, not just a privileged few.

The jobs of the future will be created by those who understand that the most potent technology in any economy is not the algorithm, but the human capacity to learn, adapt, and innovate. Our policy focus must be singular: to empower that capacity at scale.

 

The Fed’s High-Stakes Swan Song: Managing the Energy Mirage

 

The Fed’s High-Stakes Swan Song:  Managing the Energy Mirage

R Kannan

In the hallowed halls of the Eccles Building, the mood this week was not one of decisive action, but of studied, perhaps even anxious, deliberation. The Federal Open Market Committee’s (FOMC) decision to maintain the federal funds rate at 3.50% to 3.75%—a move widely telegraphed but nonetheless weighty—underscores the precarious tightrope walk facing American monetary policy. As Chair Jerome Powell conducted what may well be his final press conference, the committee’s message was clear: the Federal Reserve is not merely waiting for data; it is waiting for clarity in a fog of geopolitical and supply-side complexity.

 

For the casual observer, the decision to hold rates steady might look like passivity. Yet, a deeper reading of the minutes and the accompanying commentary reveals a fractured consensus. Three hawkish dissents on the forward guidance language serve as a flashing warning light, signalling that the unified front the Fed has projected for years is beginning to crack under the pressure of divergent economic theories and mounting uncertainty.

The central challenge, as articulated by the Fed, is a classic monetary paradox. We are currently witnessing an inflation profile that is undeniably elevated, driven significantly by a spike in global energy prices. Traditional macroeconomic doctrine dictates that when inflation remains sticky, the central bank must tighten the screws to dampen demand. However, the Fed is acutely aware that these same energy prices are functioning as a "stealth tax" on the American consumer. By increasing fuel and heating costs, this inflation shock is already actively cooling the economy, acting as a natural, albeit painful, brake on discretionary spending.

In this light, the Fed is trapped. To tighten policy further to combat the inflationary "shock" would be to risk over-correcting, potentially pushing a cooling economy into a needless contraction. To signal easing would be to risk unmooring inflation expectations at a time when the "back side" of the energy shock remains invisible. Thus, we are left with the "wait-and-see" posture—a stance that is academically defensible but increasingly risky in practice.

The institutional subtext of this meeting cannot be ignored. Chair Powell’s defence of the Fed’s independence and his commitment to "transparency and finality" regarding the ongoing legal and political pressures surrounding the institution felt like a closing argument. As the Fed prepares for a significant leadership transition in the coming months, the uncertainty surrounding who will helm the world’s most powerful central bank is beginning to bleed into market sentiment. When leadership is in flux, the temptation is often to default to the status quo. However, the American economy in mid-2026 is not a static environment. It is a dynamic system reacting to geopolitical volatility, shifting labour dynamics, and the lagging effects of previous policy adjustments.

The real danger in the current outlook is not just inflation or recession; it is policy obsolescence. If the Fed remains wedded to a data-dependent strategy that relies on lagging indicators while the structural underpinnings of the economy are shifting rapidly due to the energy crisis and geopolitical realignment, they risk fighting the last war. The labour market, while showing resilience, is beginning to fray at the edges, and the cooling of consumer confidence suggests that the "soft landing" narrative is becoming harder to justify.

Looking ahead, the next few months will be a crucible for the institution. If the energy shock persists, the Fed will have to confront the reality that its dual mandate—maximum employment and price stability—is increasingly in conflict. We can no longer assume that a cooling economy will automatically be cured by lower energy prices, nor that inflation will dissipate without more aggressive intervention.

For the American economy, the outlook for 2026 remains cautiously pessimistic. We are moving toward a period where "no news is bad news." Stagnant policy in the face of dynamic global challenges is effectively an admission that the Fed has run out of easy levers. As the committee waits for the "back side" of the energy spike, businesses and households are left to navigate a high-interest-rate environment that is increasingly disconnected from the reality of tightening margins and slowing growth.

The Fed’s swan song under Powell is a reminder that central banking is not a science; it is a precarious art. By choosing to hold steady, the committee has bought itself time, but it has not bought itself a solution. The transition in leadership will be the ultimate test of the institution’s durability. For the American public, the hope must be that the next chapter of Fed policy offers more than just a continuation of the current, agonizing equilibrium. We need a central bank that is not just attentive to the risks on both sides of its mandate, but one that is willing to define a path forward that recognizes the world as it is today, not as we hope it will be tomorrow.

 

 

The Great Fragmentation: Mapping the New Contours of Global Trade

 The Great Fragmentation: Mapping the New Contours of Global Trade

R Kannan

For nearly three decades after the fall of the Berlin Wall, the narrative of global trade was one of relentless, borderless integration. The "End of History" was supposed to be paved with container ships, low tariffs, and the hyper-efficiency of just-in-time supply chains. Today, that world is unravelling. In its place, a more fractured, securitized, and complex landscape is emerging—what economists at the International Monetary Fund (IMF) and the World Bank are increasingly labelling "Gated Globalization."

 

From the financial hubs of Mumbai to the volatile shipping lanes of the Red Sea, the signals are clear: the era of efficiency-first trade is being replaced by an era of security-first trade. According to the latest reports from the World Trade Organization (WTO) and the United Nations, global trade is undergoing its most profound structural shift since the founding of the General Agreement on Tariffs and Trade (GATT) in 1947.

The Rise of "Geoeconomic Fragmentation"

The primary driver of this shift is the increasing weaponization of trade policy for geopolitical ends. In its World Economic Outlook (April 2026), the IMF warns that "geoeconomic fragmentation" is no longer a theoretical risk but a present reality. US effective tariff rates, which sat at roughly 2.4% in late 2024, surged to 15% by the end of 2025—the highest levels since the post-World War II reconstruction era.

This is not merely a bilateral dispute between the U.S. and China. Fragmentation is spreading across the G20 and beyond. The European Union has implemented new "strategic autonomy" safeguards on steel and chemicals, while Mexico recently introduced surcharges of up to 50% on a range of imports to protect domestic industries from perceived dumping. The Wall Street Journal reports that trade policy is now being "shaped by security and political considerations rather than efficiency or multilateral rules," leading to a world where trade blocks are increasingly insular.

From Offshoring to "Friend-Shoring"

The most visible trend in this new era is the death of the traditional "offshoring" model. During the "hyper-globalization" phase (2002–2007), companies moved production to wherever labour and capital costs were lowest. Today, the focus has shifted to "Resilience" and "De-risking."

UNCTAD’s 2025 reports highlight a sharp resurgence in "Friend-shoring"—the practice of restructuring supply chains to favour trade with politically aligned partners. This trend is particularly pronounced in strategic sectors such as semiconductors, electric vehicles (EVs), and critical minerals. In these industries, countries are prioritizing "technological sovereignty" over pure cost-efficiency.

As a result, we are seeing the emergence of new regional hubs. While US imports from China have dropped sharply in relative terms, countries like Vietnam, Taiwan, and Mexico have seen a surge in trade volume. However, the IMF cautions that this is often "indirect trade." Many goods are still manufactured with Chinese components and merely assembled in "friendly" third countries, creating a more opaque, more expensive, and potentially more fragile version of the old global supply chain.

The Digital Paradox: Services in an Age of Barriers

While trade in physical goods faces significant headwinds, digital trade is moving in the opposite direction. The WTO’s World Trade Report 2024 emphasizes that digitally delivered services—ranging from streaming and software to remote professional services and AI architecture—are the fastest-growing segment of global trade.

This "Digital Paradox" suggests that while it is becoming harder to ship a car or a turbine across a border due to physical and regulatory hurdles, it is becoming easier to ship the software that runs them. UNCTAD estimates that growth in digital services trade will continue to outpace goods trade through 2026. However, a new threat looms: data localization laws. The Financial Times notes that if data is treated as a "national asset" that cannot leave borders, digital trade could soon face its own version of the high tariffs currently hitting the manufacturing sector.

The Green Trade Revolution and Carbon Protectionism

Climate change is also rewriting the rules of the game. The "Green Transition" is fostering a new, more sophisticated type of protectionism. Governments are increasingly using massive subsidies and "carbon border adjustment mechanisms" (CBAMs) to protect domestic green industries while penalizing carbon-intensive imports.

The World Bank’s Trade Fragmentation Research Initiative notes that while these policies aim to reduce global emissions, they often create uncoordinated trade barriers that disproportionately hurt low-income economies. Developing nations, many of which are commodity-dependent, face heightened price volatility as they struggle to adapt to the rigorous green standards imposed by advanced economies like the EU. This "Green Squeeze" is becoming a central point of contention in North-South trade relations.

The Role of Financial Stability and Gold

As the trade landscape fragments, the financial foundations of global commerce are also shifting. The New York Times reports a significant increase in central bank gold purchases, particularly in emerging markets, as a hedge against a weakening or "weaponized" US dollar.

The volatility of the dollar, combined with the rise of regional currencies in trade settlements (such as the "petro-yuan" or local currency settlement systems in ASEAN and BRICS+), is complicating the traditional "dollars-for-goods" model. The IMF warns that a multi-currency trade world, while potentially more diverse, carries higher transaction costs and greater exchange rate risks for small-to-medium enterprises.

Re-Globalization vs. De-Globalization: The Path Forward

Despite the prevailing gloom, the WTO argues that we are not witnessing the end of globalization, but its "re-globalization." The World Trade Report 2024 makes a passionate case that trade remains the most effective tool for income convergence and poverty reduction. The challenge, according to the UN’s World Economic Situation and Prospects, is that the benefits of trade are currently being concentrated among a few "aligned" blocks, leaving the most vulnerable nations behind.

Reforming the dispute settlement mechanism—which has been paralyzed for years—and addressing the specific needs of the Global South will be critical to preventing a total collapse of the rules-based order.

Conclusion: A World of "Episodic Shocks"

As we move toward 2027, the global economy appears to have entered a period where "fragility and episodic shocks are increasingly structural features," per the IMF. For global corporations and national governments, the strategy is no longer about maximizing growth at all costs, but about managing risk in a world that is less coordinated and more risk-averse.

The "Great Convergence" that defined the early 21st century has stalled. In its place, we find a world of "strategic power gaps" being filled by regional alliances and protective walls. Global trade is not dying, but it is becoming a much more expensive and complicated game to play. The winners in this new era will not be those with the lowest costs, but those with the most resilient and politically astute supply networks.

 

The Mythos of Security: Why AI-Driven Exploitation Demands a "Biological" Defence

The Mythos of Security: Why AI-Driven Exploitation Demands a "Biological" Defence

By R. Kannan

The traditional perimeter of global enterprise has not just been breached; it has been rendered obsolete. In April 2026, the release of frontier models like Anthropic’s Claude Mythos signalled a permanent shift in the balance of power between the digital lock and the digital pick. We have entered the era of autonomous exploitation, where software vulnerabilities—some lying dormant for nearly three decades—are being unearthed and weaponized in minutes by machine intelligence.

For the modern CEO and the boards they report to, the message is chilling: the window of opportunity for human-led defence has shrunk from months to mere seconds. If our defensive posture remains anchored in human reaction times and periodic audits, we are essentially fighting a supersonic war with a cavalry mindset.

 

 

To address the exponential threat posed by autonomous exploitation models like Claude Mythos, defensive strategies must evolve from static checklists to dynamic, machine-speed ecosystems.

What to do

I. Strategic Infrastructure & Governance

Establish an AI Threat War Room

A traditional Security Operations Centre (SOC) is reactive, often mired in "alert fatigue." The AI Threat War Room is a proactive command centre staffed by "Purple Teams"—specialists who blend offensive (Red) and defensive (Blue) tactics.

• Offensive Synthesis: The team utilizes adversarial AI to simulate multi-stage attacks. This involves "LLM-orchestrated" fuzzing, where the AI generates millions of permutations of inputs to break your proprietary software.
• Predictive Remediation: Instead of waiting for a CVE (Common Vulnerabilities and Exposures) to be published, this unit identifies "silent" weaknesses in logic and business workflows that traditional scanners miss.
• Executive Oversight: This room provides the Board with a real-time "Resilience Scorecard," translating technical vulnerabilities into enterprise risk metrics.

Zero-Trust Architecture (ZTA)

The "Castle and Moat" philosophy is dead. ZTA operates on the mantra: "Never Trust, Always Verify."

• Identity-as-the-New-Perimeter: Access is not granted based on being "on the office Wi-Fi." Every request—from a CEO's laptop or a cloud microservice—requires cryptographic verification and device health attestation.
• Contextual Risk Engines: ZTA uses AI to analyse the "signals" of a login. If a user logs in from Mumbai but their device lacks the latest security patch, or the typing cadence (biometrics) doesn't match, access is denied or "stepped up" to higher authentication.
• Least Privilege Enforcement: Users only see the applications necessary for their specific role. This "darkens" the rest of the network to a potential attacker.

Aggressive "Technical Debt" Liquidation

Legacy systems (Mainframes, old Windows servers, unpatched ERPs) are "sitting ducks" for AI like Mythos, which can scan decades-old codebases in seconds.

• Vulnerability Aging Analytics: Categorize all software by its "exploitability age." Any system running code that hasn't been refactored in 5+ years should be moved to an "Isolated Legacy Zone."
• The "Sunsetting" Mandate: Establish a rigid policy where "End-of-Life" (EOL) means immediate disconnection. If a business unit requires an EOL tool, they must pay a "Security Tax" to fund its modernization.
• Cloud-Native Migration: Prioritize moving legacy workloads to "Serverless" or "Containerized" environments where the underlying infrastructure is patched automatically by the cloud provider.

Micro-Segmentation

In a flat network, one compromised password leads to a total data breach. Micro-segmentation creates "digital bulkheads" similar to a submarine.

• Application-Level Isolation: Every application is wrapped in its own micro-perimeter. A breach in the "Marketing Analytics" tool cannot jump to the "Payroll Database."
• Dynamic Policy Generation: Using AI to observe traffic patterns, the system automatically drafts firewall rules that allow only necessary communication (e.g., "Web Server A can only talk to Database B on Port 443").
• Blast Radius Limitation: Even if an AI agent gains "Admin" rights within one segment, it finds itself trapped in a "cell," unable to see or reach other critical enterprise assets.

Software Bill of Materials (SBOM)

Modern software is a "Lego set" of third-party libraries. If one small library (like Log4j) is vulnerable, your entire enterprise is at risk.

• Supply Chain Transparency: Demand a machine-readable SBOM (in formats like CycloneDX) from every software vendor. This is essentially a "list of ingredients."
• Real-Time Dependency Mapping: If an AI model discovers a zero-day in an obscure open-source library, your SBOM system should instantly flag every application in your company that uses it.
• VEX (Vulnerability Exploitability eXchange): Integrate SBOMs with VEX data to determine not just if a "vulnerable library" exists, but if the library is actually "reachable" and "exploitable" in your specific configuration.

II. AI-Native Defence Operations

Deploy "Virtual Patching"

The "Vulnerability-to-Patch" gap is where hackers win. It takes humans weeks to test and deploy a patch; AI exploits the bug in minutes.

• Immediate Shielding: When a vulnerability is identified, a Web Application Firewall (WAF) or an Intrusion Prevention System (IPS) applies a "virtual patch"—a rule that specifically blocks the traffic pattern required to exploit that bug.
• Zero-Downtime Security: This allows the company to stay protected without rebooting critical servers or disrupting business operations while developers work on the permanent code fix.
• Automated Signature Generation: Advanced defence tools can now analyse a new exploit and write their own virtual patch rules in milliseconds.

Automated Red Teaming

Security is no longer a "once-a-year" audit. It is a continuous battle.

• Continuous Adversarial Simulation: Deploy "Defensive AI" agents that act as "Chaos Monkeys." They constantly try to trick your employees with AI-generated phishing, probe your cloud buckets for misconfigurations, and attempt to crack passwords.
• Evidence-Based Security: Instead of wondering "Are we secure?", you have a daily report of exactly which attacks were attempted and which ones were stopped.
• Evolving Defence: As the Red Team AI learns new tricks from global threat intelligence, your Blue Team (defenders) automatically receives updates on how to counter those specific techniques.

Agentic SOC Orchestration

The human brain cannot process 100,000 security alerts per day. Agentic AI can.

• Reasoning-Capable Agents: Unlike old automation (which followed "If-This-Then-That" rules), Agentic AI can "think." It can see an alert, decide to look at the user's recent emails, check the server logs, and determine if the activity is a real attack or a false alarm.
• Automated Remediation: If a breach is confirmed, the AI agent can autonomously isolate the infected laptop, reset the user's password, and notify the legal team—all in under 30 seconds.
• Cross-Tool Intelligence: These agents act as a "connective tissue" between your firewall, your email security, and your cloud logs, creating a unified defence narrative.

Outbound Traffic Filtering (Egress Control)

Most security focuses on who is entering the network. In the age of data theft, who is leaving is more important.

• "Default Deny" for Outbound: Production servers should never be able to browse the general internet. They should only be allowed to talk to specific, pre-approved update sites or APIs.
• Command & Control (C2) Blocking: When an AI agent infects a system, it must "call home" to receive instructions. Rigorous outbound filtering breaks this link, rendering the malware "blind and deaf."
• Data Exfiltration Prevention: Use AI to monitor the volume and destination of outgoing data. A sudden 50GB transfer to an unknown IP address in a foreign country should be blocked instantly.

Behavioural Anomaly Detection

Hackers today don't "break in," they "log in" using stolen or AI-guessed credentials.

• User & Entity Behaviour Analytics (UEBA): Establish a "baseline of normal" for every employee. If a Corporate Advisor who usually reads "Strategic Reports" suddenly starts downloading "SQL Database Schemas," the system flags the behaviour as an anomaly.
• Time & Velocity Checks: If an account logs in from Mumbai at 9:00 AM and from London at 9:05 AM, the system detects "impossible travel" and locks the account.
• Process Integrity: Monitor how software behaves. If the "Calculator" app suddenly tries to access the "Microphone" or the "Keychain," the AI defence identifies this as a "Process Injection" attack and kills the task.

 

Expert Insight for the Board: The transition to these  steps requires a cultural shift from "Security as a Cost Centre" to "Cyber-Resilience as a Competitive Advantage." In 2026, the companies that survive Claude Mythos-style attacks will be those that treat their digital infrastructure as a living, self-healing organism.

To combat the speed of Claude Mythos, your Identity, Supply Chain, and Recovery protocols must shift from "static barriers" to "dynamic ecosystems."

III. Identity & Access Management (IAM)

Just-in-Time (JIT) Privileges

In a traditional setup, an admin has "god-mode" keys 24/7. If an AI compromises that account at 2 AM, it’s game over. JIT turns these into "Cinderella permissions."

• Ephemeral Tokens: Access is granted via a temporary token that expires in 30, 60, or 120 minutes. Once the task is done, the "key" dissolves.
• Approval Workflows: For high-risk systems, the AI defensive layer requires a "second set of eyes" (human or a verified secondary AI) to authorize the elevation of privileges.
• Zero Standing Risk: By ensuring no one has permanent admin rights, you remove the most valuable target from the attacker’s map. Even if a password is stolen, it grants zero power until a JIT request is validated.

Non-Human Identity (NHI) Governance

Modern enterprises have 10x more "bot" identities (API keys, service accounts, secrets) than human ones. Mythos targets these because they rarely have MFA.

• Secret Rotation: Automatically rotate API keys and passwords every 24 hours. This shrinks the "usability window" for a stolen credential.
• Scoped Permissions: Ensure a service account meant to "Read Weather Data" doesn't have the permission to "Delete Database."
• NHI Discovery: Use AI to find "orphaned" accounts—old bots left behind by former developers that still have access to production environments.

Phishing-Resistant MFA

Traditional 2FA (SMS or App Push) is now trivial for AI to bypass via "MFA Fatigue" attacks or proxy-phishing sites.

• FIDO2 / WebAuthn: Shift to hardware keys (YubiKeys) or device-level Passkeys. These use asymmetric cryptography; the secret never leaves the hardware, making it impossible for an AI to "intercept" the code.
• Eliminating the "Human Hook": By removing the need for a user to type a 6-digit code, you remove the opportunity for an AI to trick them into typing that code into a fake website.

Contractor Credential Hardening

External partners are the "Trojan Horse" of the modern enterprise.

• Siloed Environments: Contractors should work in isolated Virtual Desktop Infrastructures (VDI). They see a screen, but the data never actually touches their local machine.
• Time-Bound Access: Contractor accounts should automatically disable themselves every Friday evening and require re-validation every Monday morning.
• Monitoring "Normalcy": If a contractor’s account usually accesses three specific folders but suddenly starts scanning the entire network, the AI defence should terminate the session instantly.

IV. Development & Supply Chain Security

AI-Integrated CI/CD Pipelines

If your developers are using AI to write code, your security must use AI to check it.

• Static & Dynamic Analysis (SAST/DAST): Integrate "Guardrail AI" into the deployment pipeline. If code contains a logic flaw that Mythos could exploit, the build is "broken" and cannot be deployed to the cloud.
• AI Code Review: Use Large Language Models trained specifically on cybersecurity to read pull requests, flagging not just syntax errors but "semantic vulnerabilities" (e.g., insecure handling of user data).

Managed Artifact Repositories

The "Open Source" world is a minefield of poisoned packages.

• Quarantine Zones: All new libraries downloaded from the internet must sit in a "quarantine repository" for 24 hours while an AI red-teams them for hidden backdoors.
• Version Pinning: Never use the "latest" version of a tool automatically. Use a verified version that has been vetted by your internal security team.
• Digital Signatures: Ensure every piece of code used in your production environment is digitally signed, proving it hasn't been tampered with since it was vetted.

SaaS Posture Management (SSPM)

A single "Public" checkbox in a Salesforce or S3 bucket can leak millions of records.

• Configuration Drift Detection: AI constantly compares your current SaaS settings against a "Golden Standard." If a user accidentally makes a Slack channel public, the SSPM tool switches it back to private automatically.
• Cross-Platform Visibility: Get a single dashboard that shows the security health of Microsoft 365, AWS, ServiceNow, and Zoom simultaneously.

Data Loss Prevention (DLP) for GenAI

Employees often "leak" secrets by asking public AI models to "debug this code" or "summarize this confidential meeting."

• AI Firewalls: Intercept prompts sent to public LLMs. If the prompt contains a credit card number, a private API key, or internal IP addresses, the system redacts the data before it leaves the company.
• Enterprise AI Tunnels: Provide employees with internal, "sanitized" versions of AI tools (like a private instance of Claude or ChatGPT) where the data stays within your corporate boundary and is not used for training.

V. Resilience & Recovery

Immutable Backups

Ransomware now targets backups first to ensure you have to pay.

• WORM Storage: Use "Write Once, Read Many" technology. Once data is backed up, it physically cannot be modified or deleted by any user (even an admin) for a set period (e.g., 30 days).
• Air-Gapped Copies: Keep one copy of your most critical data entirely offline. If the cloud is compromised, the "Gold Copy" remains untouched.
• Automated Recovery Testing: Use AI to constantly "rehearse" the recovery of your data. If a backup is corrupted, you need to know before the disaster strikes.

AI-Specific Tabletop Exercises

Traditional disaster drills are too slow. You need "War Games" for the AI era.

• Hyper-Speed Simulations: Run drills where the "attack" happens in real-time. Can your team make a decision in 2 minutes? If not, what parts of the decision-making process can be handed over to an AI agent?
• The "Human-in-the-Loop" Test: Determine exactly where a human must be involved and where they are just a bottleneck.
• Psychological Readiness: Train staff to recognize "Deepfake" audio or video from the CEO asking for emergency fund transfers or password resets—a hallmark of Mythos-era social engineering.

The New Bottom Line: MTTR vs. MTTD

In the past, we focused on Mean Time to Detection (MTTD)—how long until we see them? In the era of Claude Mythos, detection is instant because the AI is loud and fast. The only metric that matters now is Mean Time to Remediation (MTTR).

Conclusion

The release of Claude Mythos is a "Sputnik moment" for global enterprise. It has exposed the fragility of the digital foundations upon which the global economy is built. However, this is not a counsel of despair. It is a call for an evolutionary leap.

By adopting AI-native defence, embracing zero-trust, and focusing on the speed of remediation over the height of the wall, companies can build a new kind of resilience. We cannot stop the AI from finding the weak points, but we can build systems that are too fast, too segmented, and too "biologically" adaptive for those weak points to matter. The future belongs to the agile, the autonomous, and the resilient. The era of the "unbreakable" castle is over; the era of the self-healing organism has begun.