The Americans have decided frontier models are not global software products; they are strategic assets, as tightly locked as nuclear centrifuges or fighter jets.
There is a moment—a collision—where the abstract promise of progress hits the hard wall of politics. For artificial intelligence, it happened at 5:21 PM on June 12… an email from the US Commerce Department landing in the inbox of Anthropic, a San Francisco entity valued near a trillion dollars. The mandate was absolute: shut down Fable 5 and Mythos 5 for every foreign national, even those within their own walls. Unable to verify passports in real time, they simply turned off the lights for everyone. An Iranian researcher in California, a French engineer, a customer in Mumbai… all discovered that humanity’s most powerful reasoning tool was no longer theirs. Their crime? Not negligence, but merely the wrong citizenship.
This is the new geography of AI, and India stands at a fracture that will define the next half-century. The Americans have decided frontier models are not global software products; they are strategic assets, as tightly locked as nuclear centrifuges or fighter jets. They used the deemed export doctrine—a Cold War relic treating access for a foreign national as an export to their home country, even on American soil. For decades, this governed uranium and semiconductors… last week, it governed a chatbot. With 1.2 million international students and 3.2 million visa workers in the US—many Indian—the architects of Mythos 5 are now legally barred from its use. Anthropic’s data is clinical: offer acceptance for international researchers collapsed from 78% to 52%. The brain drain policymakers feared may finally reverse.
But what does this mean for our own technological horizon? For years, the comfortable delusion in Delhi and Mumbai boardrooms was that India could ride the American wave—using their clouds, models, and APIs to power digital transformation without the crushing cost of building the core. That assumption died on June 12. Anthropic’s blackout was not a technical glitch; it was a political decision, devoid of warning or judicial review, leaving millions who built operations around this tool in the dark. If an Indian hospital uses American AI for diagnostics, or a bank for fraud, or a government for public service, they now know with absolute certainty: their service can be extinguished by a foreign government in the time it takes to send one email.
The irony? In trying to slow China, the effect may be opposite. Chinese models like Qwen and DeepSeek trail American ones by less than 3% on benchmarks, yet cost a fraction. More importantly, they are open weights… once downloaded, they run locally, immune to the whims of Washington or Beijing. Since restrictions began, downloads of these alternatives surged 340%. Corporate users, watching a trillion-dollar flagship vanish overnight, are quietly building operational dependence on Chinese systems as insurance against American political caprice. The Financial Times was brutally clear: cutting access to Anthropic is a gift to China.
For India, this is both a warning and an opening. The warning: dependence on American AI infrastructure is no longer commercial convenience; it is strategic vulnerability. The opening: the global landscape is fracturing, leaving space for a third pole—if we have the ambition. Europe learned this painfully; the EU spent years crafting the AI Act, only to find that regulation without capability is merely the right to be governed by others’ technology. India should not need to learn this lesson the same way.
There is another dimension of national security. Mythos 5 was more than a chatbot; it was the engine for Project Glasswing, using advanced reasoning to find zero-day vulnerabilities in critical infrastructure across fifteen countries, including NATO. The ban killed it instantly. As AI cyber threats escalate, the US disabled a primary defensive tool. This is the cybersecurity paradox: in the name of security, America reduced global security. For India’s infrastructure operators, the lesson is unmistakable… defensive capabilities cannot be outsourced to platforms subject to foreign political whim.
The timing carries its own bitter irony. Anthropic’s CEO, Dario Amodei, published a policy essay calling for government authority just one day before launch; within 48 hours, the Trump administration used that authority as a weapon against him. Since last February, Anthropic has clashed with the Pentagon over autonomous weapons and mass surveillance. If advocating for AI safety invites retaliation, transparency evaporates… companies will hide risks, and the public will pay the price.
For Indian enterprises, the technical lesson arrives with immediate, clinical force. Teams who hardcoded Fable 5 into their systems faced total failure—a sudden blackout; meanwhile, those utilizing abstraction layers with automatic fallback simply redirected their traffic… a seamless pivot. Model-agnostic architecture, once viewed as a luxury, has now become a business continuity imperative, a necessary shield against caprice. Any public service dependent on a single foreign AI provider is essentially a service that can be extinguished by a foreign government in an instant. The National Informatics Centre, our state governments, and various public sector undertakings must now perceive AI access not as a utility, but as a strategic vulnerability… a fragile thread that can be severed without warning, leaving the machinery of state to fumble in the dark.
Against this backdrop, conversations in Delhi’s corridors and Bengaluru’s campuses have acquired a new urgency. For months, an Indian frontier model was a distant aspiration; the Anthropic ban collapsed that timeline. MeitY officials are now weighing if IndiaAI Mission’s infrastructure can scale to frontier levels, while IIT researchers discuss adapting open-source architectures with Indian linguistic diversity. The challenge is formidable—requiring tens of thousands of chips, billions of dollars, and years of expertise. Talent is abundant, but hardware and capital remain concentrated in American and Chinese hands. Yet the alternative is no longer abstract; it is the certainty that India’s digital future runs on infrastructure a foreign government can kill in minutes. The question is no longer if India can afford to build its own model… but if we can afford not to.
What emerges is an uncomfortable truth—a realization that citizenship is becoming the primary determinant of technological access. Your passport now dictates whether you may use the tools for building the future… a boundary drawn in ink and bureaucracy. The Americans prioritize their security calculations over global access; the Chinese build alternatives immune to such control, operating in their own orbit. India’s path must be different—neither dependent nor merely imitative, but sovereign, open, and genuinely capable. The global AI order is being rewritten in real time… a script unfolding across borders. Will we be authors of this new chapter, or merely readers of other people’s drafts? The switch has been flipped; the lights have gone out in one room, leaving us to contemplate the void. Whether we build our own illumination, or sit in darkness waiting for others to restore the power, is a choice we must make now… before the haze consumes everything.
*Brijesh Singh is a senior IPS officer and an author (@ brijeshbsingh on X). His latest book on ancient India, “The Cloud Chariot” (Penguin) is out on stands. Views are personal.
